# Applied / Comms With AI
> Grounded in hands-on experimentation, workflow transformation, and real-world implementation.
Public Ghost content for AI and LLM tooling. This file includes a bounded export of public pages first, then recent public posts.
Append `.md` to any post or page URL to get the content in Markdown (for example, `/example-post.md`).
## Pages
### About
URL: https://www.appliedcomms.ai/about/
Last updated: 2026-09-02T17:29:56.000Z
Applied / Comms With AI is the publication arm of [Comms With AI](https://www.commswith.ai/). It is written from inside the work: building the systems, running them with real organisations, and reporting what happened, including the parts that did not go to plan.
## **What you will find here**
- Long-form articles on how communications work is changing, written to be useful rather than predictive.
- Transparent experiments, where the method and the results are published together so you can judge them.
- Tool reviews based on use against real briefs.
- And the Comms With AI Leader Interview series, live conversations with the people shaping how the profession thinks about this.
## **Who writes it**
I am Michael MacLennan, a communications strategist based near Edinburgh. I have led communications at senior level for more than twenty years, including five as Director of Social Media Strategy, Digital and Comms at Brunswick Group, and earlier digital work at Red Bull and Barclaycard. My career started in newsrooms, at The Herald and STV.
In 2024 I founded [Faur](https://faur.site/), and in 2026 launched Comms With AI, which won two awards at the inaugural AI Comms Awards: Gold for Best Innovation in AI Tools for Communications, and AI Communications Leader of the Year (Agency). The judges' words were "simple, specific and scalable", which is the standard everything here is held to.
I hold a Professional Certificate in Machine Learning and Artificial Intelligence from Imperial College London, completed in 2025\. It matters for one reason: it means I can design, build and run the systems I write about, rather than describe other people's. I am also a director on the board of ScotlandIS, Scotland's digital technology industry body.
## **How the three brands fit together**
[Comms With AI](https://www.commswith.ai/) is where the templates, workflows and tools live: things a communications team can put to use in fifteen to thirty minutes. Applied / Comms With AI is where the thinking behind them is worked out in public. [Faur](https://faur.site/) is the consultancy for organisations that need senior, human-led help with AI adoption across the whole business, from use policy and governance to rollout and measurement.
Most of what gets tested here ends up in the Comms With AI library, and most of what Faur learns in client work ends up here, with the client's details removed.
## **The ground rules**
Nothing is recommended that has not been built and used. Where something did not work, that is what gets reported. Every experiment shows its method, so you can disagree with the conclusion on the evidence. The focus stays on communications: corporate, crisis, reputation, internal and public affairs. And AI-generated text is never passed off as mine; the [ethics policy](https://www.appliedcomms.ai/ethics-policy/) sets out how it is used and disclosed.
## **A note on uncertainty**
Nobody has the full picture on AI in communications, and the practitioners worth reading are the ones who say so. What this publication can offer is direct implementation experience and a commitment to showing the work. If you are after confident predictions, there is no shortage of those elsewhere.
## **Get in touch**
Find me on [LinkedIn](https://www.linkedin.com/in/michaelmaclennan/) . The publication's own page is [Applied / Comms With AI on LinkedIn](https://www.linkedin.com/showcase/appliedcommsai/).
### Contact
URL: https://www.appliedcomms.ai/contact/
Last updated: 2026-02-25T14:37:34.000Z
**Got a question, a story to share, or a project to discuss?** Here's where to go.
---
## Work with Faur
Applied Comms AI is the learning and experimentation arm of Faur – a communications consultancy specialising in AI implementation, strategy, and content.
If you're looking for hands-on support – whether that's an AI implementation project, a strategic communications workshop, or a paid consultation – Faur is where that happens.
[**Book a consultation →**](https://faur.site/services/consultation) A 45-minute session with a comprehensive advisory note. £250.
[**See all Faur services →**](https://faur.site/contact) From retained advisory to associate-led projects and team training.
---
## For Newsletter Readers
### Story submissions and case studies
I'm always looking for real-world AI implementation stories to feature in the newsletter – successes, failures, and everything in between. If you've tested something that other communications professionals should know about, I'd like to hear it.
Anonymous contributions are welcome. Just tell me what you tried, what happened, and what you'd do differently.
### Guest contributions
Applied Comms AI occasionally features guest contributions from practitioners with direct experience of AI implementation in communications. If you have something genuinely useful to share – grounded in your own work, not theoretical – get in touch.
### Feedback and suggestions
Questions about specific AI applications, tool recommendations, or suggestions for topics I should cover are all welcome. The newsletter is better for reader input, and I read everything.
**Email:** [michael@faur.site](mailto:michael@faur.site)
---
## Connect
- **LinkedIn:** [Michael MacLennan](https://www.linkedin.com/in/maclennanmichael/)
- **Applied Comms AI socials:** [LinkedIn](https://www.linkedin.com/showcase/appliedcommsai/?viewAsMember=true) / [Instagram](https://www.instagram.com/appliedcommsai/) / [Threads](https://www.threads.com/) / [Bluesky](https://bsky.app/profile/appliedcommsai.bsky.social)
- **Faur:** [faur.site](https://faur.site)
---
## About Applied Comms AI
Applied Comms AI documents real AI implementation for communications professionals. It's part of the Faur ecosystem alongside [Comms With AI](https://www.commswith.ai) – a library of ready-to-use templates, prompts, and workflows.
### Ethics Policy
URL: https://www.appliedcomms.ai/ethics-policy/
Last updated: 2026-03-27T08:05:30.000Z
## Our Commitment
Applied Comms AI believes artificial intelligence should enhance human creativity and capability in communications, not replace it. We're committed to using AI responsibly whilst honestly documenting our journey, including the sustainability challenges we're still working to solve.
**We are signatories to the** [**Global Alliance's Responsible AI Guiding Principles**](https://www.globalalliancepr.org/guiding-principles-for-ethical-and-responsible-artificial-intelligence) and have committed to the Venice Pledge, which provides our profession with a shared framework for the responsible use of AI. Our policy builds upon these seven foundational principles whilst addressing the specific challenges of our newsletter and community.
## Core Principles
*These principles align with and build upon the Global Alliance's seven Responsible AI Guiding Principles, which have been adapted for our specific newsletter and community context.*
### 1\. Transparency First
- **Always disclose** when content is AI-assisted or AI-generated
- **Share our process**: How we used AI, what worked, what didn't
- **Admit limitations**: We won't claim AI can do things it can't
- **Show the human element**: Highlight where human judgement remains essential
### 2\. Quality & Authenticity
- **AI augments, humans decide**: All final editorial decisions remain human-made
- **No AI ghostwriting**: If content is attributed to a person, that person wrote it
- **Fact-check everything**: AI outputs are always verified before publication
- **Preserve voice**: AI should enhance our tone, not replace it
### 3\. Respect for People
- **Protect privacy**: No personal data fed to AI systems without consent
- **Credit human work**: Acknowledge when AI builds on existing human creativity
- **Employment impact**: Openly discuss AI's effect on communications jobs
- **Accessibility**: Ensure AI doesn't create barriers for any audience
## Professional Standards & Industry Alignment
As signatories to the Global Alliance's Responsible AI Guiding Principles, we commit to:
- **Ethics First**: Adhering to professional ethical standards across all AI usage
- **Human-Led Governance** : Maintaining human oversight aligned with public interest
- **Personal and Organisational Responsibility**: Taking full accountability for all AI-assisted outputs
- **Awareness, Openness, and Transparency**: Clear disclosure of AI involvement in our content
- **Education and Professional Development**: Continuous learning and sharing knowledge with our community
- **Active Global Voice**: Contributing to industry standards and responsible AI advocacy
- **Human-Centred AI for the Common Good:** Ensuring our AI usage serves societal wellbeing
### Our Additional Standards
- **Client confidentiality**: Never use proprietary information in AI experiments
- **Intellectual property**: Respect copyright in all AI-generated content
- **Accuracy standards**: Maintain same fact-checking rigour regardless of content source
- **Bias awareness**: Actively monitor and address AI bias in outputs
## Sustainability Challenges
We acknowledge AI has significant environmental impacts through energy consumption and carbon emissions. Currently:
### What We're Doing
- **Minimising unnecessary usage**: Only using AI when it adds clear value
- **Choosing efficient models**: Selecting less resource-intensive options where possible
- **Documenting impact**: Tracking our AI usage to understand our footprint
- **Supporting research**: Highlighting sustainability innovations in AI
### What We're Working Towards
- **Carbon measurement**: Developing methods to track AI-related emissions
- **Offset consideration:** Exploring how to balance AI benefits against environmental costs
- **Efficiency advocacy**: Pushing for more sustainable AI development
- **Alternative approaches**: Testing lower-impact AI alternatives as they emerge
*We recognise this is an evolving challenge without perfect solutions yet. We'll update our approach as better options become available.*
## Content Guidelines
### When We Use AI
- **Research assistance**: Gathering background information and identifying trends
- **Draft enhancement**: Improving clarity, structure, or style of human-written content
- **Tool testing**: Experimenting with AI capabilities for newsletter content
- **Creative prompts**: Generating ideas that humans then develop
### When We Don't Use AI
- **Personal interviews**: All conversations remain human-to-human
- **Final decision-making**: Editorial choices, strategic recommendations, and ethical judgements
- **Sensitive topics**: Crisis communications, legal matters, or personal stories
- **Original insights**: Our analysis and opinions remain human-generated
## Transparency in Practice
### In Newsletter Content
- Clear labelling: e.g. "This section was written with AI assistance"
- Process notes: e.g. "We used ChatGPT to help structure this analysis, then rewrote in our voice"
- Honest assessments: e.g. "The AI got this wrong, here's what actually happened"
### In Tool Reviews
- Full disclosure of any commercial relationships
- Testing methodology is clearly explained
- Limitations and failures prominently featured
- User privacy implications discussed
## Community Standards
### Reader Engagement
- **Honest dialogue**: We'll discuss AI ethics openly with our community
- **Feedback welcome**: Readers can challenge our AI usage and we'll respond
- **Shared learning**: We'll feature reader experiences with AI ethics
- **No lecturing**: We're figuring this out together, not preaching from on high
### Industry Leadership
- **Set examples**: Demonstrating responsible AI usage in communications
- **Share standards**: Making this ethics policy available for others to adapt (we only ask that you let us know if you do, for tracking impact)
- **Challenge poor practice**: Calling out irresponsible AI usage in our field
- **Collaborate**: Working with others to improve industry standards
## Regular Review
This policy will be reviewed every six months and updated based on:
- **Technology developments**: New AI capabilities and limitations
- **Industry standards**: Evolving best practices in communications
- **Community feedback**: What our readers think works and doesn't
- **Environmental progress**: Improvements in AI sustainability
- **Our own learning**: Insights from our experiments and mistakes
## Questions or Concerns?
If you have questions about our AI ethics or would like to challenge any of our approaches, please [get in touch](https://www.appliedcomms.ai/contact/). We're committed to learning and improving.
---
*This policy reflects our commitment to responsible AI usage, and continuing to monitor and update our practices. It's a living document that will evolve as we learn more about both the opportunities and challenges of AI in communications.*
**Last updated:** 24 June 2025
**Next review:** December 2025
### Consultancy
URL: https://www.appliedcomms.ai/consultancy/
Last updated: 2025-12-05T17:00:23.000Z
## From Insights to Implementation
I'm [Michael MacLennan](https://www.linkedin.com/in/maclennanmichael/), and I've built Applied Comms AI to share what I'm learning as I help organisations actually implement AI in their communications work.
Every insight you read here comes from real projects I'm delivering through [Faur](https://faur.site/): from workflow audits for multinationals to custom agent development for membership bodies. I'm not theorising about AI in communications. I'm building it, testing it, and showing you what works.
## **What Makes This Different**
I've spent 15+ years in strategic communications – at Brunswick, Red Bull, and Barclaycard – combined with hands-on technical capability that most communicators don't have. I [recently completed Imperial College London's Professional Certificate in Machine Learning & Artificial Intelligence](https://certificates.emeritus.org/profile/michaelmaclennan614244/wallet), where I built production-ready AI systems, including an assistive moderation model that achieved top-quartile performance in competition.
That rare combination – strategic comms experience plus the ability to actually code, train models, and deploy AI systems – means I can bridge the gap between what your communications team needs and what AI can realistically deliver.
## **How I Work With Organisations**
**AI Workflow Audits & Strategy**
I assess where AI genuinely improves communications effectiveness—and where it's just an expensive distraction. You'll get honest recommendations grounded in what actually works, not vendor promises. Recent projects include community platform evaluations for renowned sports bodies and content strategy work for leading multinational corporations.
**Custom AI Development**
Sometimes what you need doesn't exist yet. I design and build bespoke systems: whether that's a morning monitoring agent for media tracking, strategic intelligence frameworks, or content workflow tools. I can write the code, train the models, and deploy systems that solve real communications challenges.
**Implementation Workshops**
Practical, hands-on sessions tailored to your organisation's maturity level. From prompt engineering fundamentals to advanced agent design, I train teams using examples from your actual work, not generic case studies.
**Crisis Communications AI Preparation**
Working with crisis specialists, I help teams develop AI-supported protocols, build tested prompt libraries, and establish governance frameworks so you're prepared before the pressure hits.
**Strategic Advisory**
Senior-level counsel for communications directors navigating AI adoption. Whether you're building business cases, managing stakeholder concerns, or designing transformation roadmaps, I provide perspective informed by both strategic experience and technical reality.
## **My Approach**
I'm transparent about what I know – and whenever there is something beyond my own expertise, I will know who best to bring in. I share failures alongside successes because in a field this new, the mistakes often teach more than the wins. I don't sell AI tools: I evaluate them honestly and help you implement what actually serves your organisation's needs.
All consultancy services are delivered through Faur, which I launched in 2024, but you'll be working directly with me. I occasionally bring in specialist associates for specific capabilities, but the strategic direction, technical implementation, and hands-on delivery comes from someone who's done work with global leaders at Brunswick and Grayling, built expansive international platforms at Red Bull, and is now building production-ready AI systems for communications teams.
## **Get in Touch**
[Explore Faur's full services](https://faur.site/services) | Email me directly: [michael@faur.site](mailto:michael@faur.site)
Currently based near Edinburgh, Scotland | Working with organisations across the UK and around the world
### The complete workflow for building campaign packs using Claude Projects and Skills
URL: https://www.appliedcomms.ai/campaign-workflow/
Last updated: 2026-09-07T10:15:49.000Z
## What This Is
This is the complete workflow from my Data Lab session on 18 February 2026, where I demonstrated building a full campaign pack in under 30 minutes using Claude Projects and Custom Skills.
Everything you need to replicate this workflow is here: the methodology, all five prompts, the setup instructions, and the principles that make it work.
## Why This Matters
Most AI workflows in communications produce one-off outputs. You ask for something, you get something, then you start from scratch next time.
This workflow is different. It builds a persistent workspace that gets more valuable with each use. Projects act as containers, Skills encode your quality standards, and each artefact you create becomes reference material for the next piece of work.
**The promise: if you set this up properly once, you'll have a reusable system that produces consistently high-quality campaign materials in 30-35 minutes.**
## The 30/70 Principle
Before we get into mechanics, understand this: AI handles roughly 30% of the work. The other 70% is your strategic judgement, editorial control, and quality assurance.
AI is excellent at structure, first drafts, and executing defined patterns. It cannot replace your understanding of stakeholders, your feel for timing, or your judgement about what's credible versus what's overreach.
Treat AI as a very capable junior strategist who follows instructions well but needs supervision. The governance pass (Step 5) exists because AI will confidently generate plausible nonsense if you don't check its work.
## The Three-Layer Framework
This workflow operates on three layers:
- **Layer 1: The Project Folder**
The persistent container for everything. Lives in your Claude account. Maintains context across multiple conversations. Holds your brief, your templates, your reference materials.
- **Layer 2: The Custom Skill**
Your quality encoder. Captures tone of voice, brand rules, writing style. Applies automatically to every generation in the Project. More reliable than repeating instructions in prompts.
- **Layer 3: The Prompts**
Execution instructions. Tells Claude what to create, in what format, using what structure. Combines with the Project's context and the Skill's rules to produce output.
*All three layers work together. Remove any one and quality drops significantly.*
## Sign up for Applied Comms AI
The practical guide for communications leaders navigating AI
Subscribe
Email sent! Check your inbox to complete your signup.
No spam. Unsubscribe anytime.
---
## Prerequisites
**You'll need:**
- Claude Pro or Team account (Projects aren't available on the free tier)
- A campaign brief (150-300 words is ideal)
- 35-40 minutes of uninterrupted time for your first run-through
- Basic understanding of what makes good comms strategy
**You don't need:**
- Deep AI expertise
- Prompt engineering experience
- Technical skills beyond copy-paste
---
## The Five-Step Workflow
### Step 1: Clean Brief (3-4 minutes)
**Goal:** Surface hidden assumptions and create a Definition of Done.
**Why this matters:** Most briefs are incomplete. Clients assume you know their sector, their constraints, their success metrics. This step forces everything into the open before you start building.
**The prompt:**
```
I'm going to paste a campaign brief. Your job is to:
1. Identify all unstated assumptions in this brief (audience demographics, budget constraints, timeline expectations, approval processes, competitive context)
2. Flag any contradictions or ambiguities
3. Propose a "Definition of Done" — the 3-5 specific deliverables and success criteria that would fulfil this brief
4. Present this as: Assumptions [in brackets], Contradictions (if any), and Suggested Definition of Done
Brief:
[PASTE BRIEF HERE]
```
**What you're looking for:**
- Assumptions flagged with \[BRACKETS\] that you can confirm or correct
- Contradictions that need resolving before you proceed
- A clear Definition of Done you can use to evaluate the final output
**Real example output:**
> "Assumptions: \[Mid-market finance teams = 50-500 employees\], \[Demo bookings = primary KPI, not MQLs\], \[2-week timeline = launch campaign, not nurture sequence\]. Definition of Done: 3 LinkedIn posts, 1 email sequence, 1 landing page copy, 1 competitor comparison guide."
**Common mistakes:**
- Skipping this step because the brief "seems clear enough"
- Not correcting wrong assumptions (AI will build the campaign on false foundations)
- Accepting vague Definitions of Done ("raise awareness" isn't measurable)
**Time check:** If this step takes more than 5 minutes, your brief is genuinely incomplete and needs more work before you proceed.
---
### Step 2: Foundational Comms Structure (5-6 minutes)
**Goal:** Create the strategic anchor for all downstream decisions.
**Why this matters:** Without a foundational structure, every subsequent piece of work will drift. This step creates a single reference document that defines key messages, audience insights, proof points, and tone rules. Everything else builds from this.
**The prompt:**
```
Create a Foundational Comms Structure for this campaign. This is the strategic anchor for all content decisions. Include:
1. Strategic Context (1 paragraph: what we're doing and why)
2. Primary Audience (demographics, psychographics, current beliefs about the category/problem)
3. Secondary Audiences (if relevant)
4. Key Message (the single most important thing this audience should understand or believe)
5. Supporting Messages (3-4 secondary messages that reinforce the key message)
6. Proof Points (specific evidence, data, or examples that substantiate each message)
7. Tone & Voice Rules (how we sound: 4-5 specific guidelines)
8. What We're NOT Saying (important boundaries or messages to avoid)
9. Success Metrics (how we'll know this worked)
Format this as a reference document. We'll return to this throughout the campaign build.
```
**What you're looking for:**
- A key message that's specific enough to guide decisions (not "we're experts" but "we're the only firm that's completed 12+ tech M&A deals under £50m in Scotland")
- Proof points that are concrete (numbers, names, verifiable facts)
- Tone rules that could differentiate two drafts ("conversational but precise" not "professional and engaging")
- Boundaries that prevent overreach ("don't claim we're disruptive — we're specialists")
**Real example output structure:**
**Key Message:**
"For compliance teams drowning in manual reporting, \[Product\] automates what currently takes 2-3 days per month into 20 minutes."
**Proof Point 1:**
Current users report 87% time reduction in compliance reporting (Q4 2025 user survey, n=47)
**Tone Rules:**
- Lead with the pain point, not the product
- Use "you" language to make it personal
- Avoid compliance jargon (no "regulatory frameworks" or "governance matrices")
- Sound relieved, not evangelical
**Critical note:** Save this document. You'll reference it constantly. If using Claude Projects, this automatically stays in the conversation history. If you're not in a Project, export it as a separate file and upload it to future conversations.
**Common mistakes:**
- Key messages that sound like mission statements ("We help businesses succeed")
- Proof points that are vague ("industry-leading results")
- Tone rules that every brand could claim ("authentic and trustworthy")
**Time check:** This should take 4-6 minutes to generate and review. If it's taking longer, the brief isn't clear enough (return to Step 1).
---
### Step 2.5: Create the Custom Skill (Optional but Recommended, 4-5 minutes)
**Goal:** Encode tone of voice rules so you don't have to repeat them in every prompt.
**Why this matters:** Custom Skills apply automatically to everything generated in the Project. Instead of remembering to paste "use UK English, write conversationally, avoid jargon" into every prompt, the Skill enforces it automatically.
**Note:** Claude's Skill builder can be unreliable. If it glitches (known issue), your tone rules in the Foundational Structure still work — you just lose the automatic application benefit.
**The setup prompt:**
```
Based on the Tone & Voice Rules in the Foundational Comms Structure, suggest 5-6 voice model options for this campaign. Each suggestion should include:
- A 2-3 word label (e.g., "Plain-speaking advisor", "Sharp-but-warm expert")
- A brief description of the voice
- A sample sentence in that voice addressing the primary audience
I'll pick one and we'll create a Custom Skill from it.
```
**What you're looking for:**
- Options that genuinely sound different from each other
- Sample sentences that demonstrate the voice (not just describe it)
- A voice that matches your client's positioning (not your personal preference)
**Example output:**
**Option 1: "Straight-talking peer"**
Sounds like: Someone who's done your job and won't waste your time.
Sample: "Compliance reporting eats 3 days a month. We've built something that gets it down to 20 minutes."
**Option 2: "Data-first advisor"**
Sounds like: Consultative but grounded in evidence.
Sample: "Our Q4 survey of 47 compliance teams showed a consistent pattern: manual reporting takes 2-3 days monthly."
**Creating the Skill (if the tool works):**
1. Click "Create Custom Skill" in Claude Projects
2. Paste the voice model description and sample
3. Add specific rules from your Foundational Structure
4. Test it by generating a short piece of content
5. Refine if it's not matching the voice you want
**If the Skill builder glitches:**
Say to your audience: "Claude's Skill builder is having a moment. The tone rules are still in our Foundational Structure, so every prompt will reference those. The Skill just saves us from having to manually include them each time — but the content quality won't suffer without it."
Then proceed with the workflow. Your tone rules are already captured in the Foundational Structure and you can reference them manually in subsequent prompts.
**Time check:** 4-5 minutes if creation works. 60 seconds to acknowledge and move on if it doesn't.
Get production-ready templates for this workflow at CommsWith.AI — tested prompts, Custom Skills, and checklists you can use immediately.
[Learn more ](https://www.commswith.ai/)
---
### Step 3: Channel Plan + Content Angles (5-6 minutes)
**Goal:** Decide where content lives and what hooks make it compelling.
**Why this matters:** Without a channel plan, you'll create content in a vacuum. This step forces strategic thinking: which channels actually reach the audience, what formats suit each channel, what angles make the content worth engaging with.
**The prompt:**
```
Based on the Foundational Comms Structure, create:
1. Channel Plan: Recommend 3-4 channels for this campaign with rationale (why these channels for this audience + goal). Consider owned, earned, and paid. Be specific (not "social media" but "LinkedIn organic posts targeting Heads of Risk").
2. Content Angles: Generate 8 distinct content hooks that would work across these channels. Each angle should:
- Connect to a supporting message or proof point
- Be specific enough to write from
- Vary in approach (some stats-led, some story-led, some problem-focused)
3. 3-Month Roadmap: Suggest a realistic content cadence and sequencing for a 3-month campaign.
Format as: Channel Plan (with rationale), Content Angles (numbered list), 3-Month Roadmap (month-by-month).
```
**What you're looking for:**
**Channel Plan:**
- Rationale that connects to audience behaviour (not "LinkedIn because B2B" but "LinkedIn because 73% of Heads of Risk follow industry thought leaders there")
- Specificity about formats (LinkedIn carousel posts, not just "LinkedIn")
- Owned/earned/paid mix (if budget allows)
**Content Angles:**
- Variety in approach (not 8 variations of "we save you time")
- Connection to specific proof points from the Foundational Structure
- Angles that could work across multiple formats
**Example output:**
**Content Angle 3:**
"The hidden cost of manual compliance: 3 days of senior time × 12 months = 36 days yearly. What could your team do with an extra month?"
*Connects to: Proof Point 1 (87% time reduction). Works as: LinkedIn post, email subject line, landing page hero.*
**3-Month Roadmap:**
- Month 1 (Awareness): Problem-focused content, competitor comparison, "day in the life" of manual compliance
- Month 2 (Consideration): Product education, demo clips, testimonial case study
- Month 3 (Decision): ROI calculator, implementation timeline, Q&A with users
**Common mistakes:**
- Channel plans based on where you want to be, not where your audience is
- Content angles that are all the same shape (8 variations of "here's why we're great")
- Roadmaps that ignore natural decision-making timelines
**Time check:** 5-6 minutes. If it's taking longer, your Foundational Structure isn't specific enough.
## Sign up for Applied Comms AI
The practical guide for communications leaders navigating AI
Subscribe
Email sent! Check your inbox to complete your signup.
No spam. Unsubscribe anytime.
---
### Step 4: Draft Assets (8-10 minutes)
**Goal:** Generate the actual content deliverables.
**Why this matters:** This is where the strategy becomes tangible. You're creating the posts, emails, landing page copy, and supporting materials that will actually go live.
**The prompt:**
```
Using the Foundational Comms Structure, Channel Plan, and Content Angles, draft 6 content assets:
1. [Specify asset type, e.g., "LinkedIn post using Content Angle 3"]
2. [Specify asset type, e.g., "Email sequence: 3 emails for demo booking nurture"]
3. [Specify asset type, e.g., "Landing page: hero, 3 benefits, CTA"]
4. [Specify asset type, e.g., "Competitor comparison: 1-page guide"]
5. [Specify asset type, e.g., "Testimonial structure for user interview"]
6. [Leave this open — let Claude suggest an appropriate 6th asset based on the campaign needs]
For each asset:
- Reference the Foundational Structure for key messages and tone
- Include [BRACKETS] for any placeholder content that needs client input
- Flag assumptions you're making about format or length
- Ensure consistency across all assets
For asset 6: Propose what would be most useful for this campaign, then draft it.
```
**What you're looking for:**
- Consistency in voice across all assets (they should sound like they're from the same campaign)
- Strategic coherence (the LinkedIn post should drive to the landing page, the email should reference the guide)
- Appropriate length for channel (LinkedIn posts = 150-200 words, not 500)
- Clear \[PLACEHOLDER\] markers for content you can't generate (client quotes, specific product features)
**Real example output (Asset 1 — LinkedIn post):**
> "Compliance reporting shouldn't take 3 days a month.
>
> For most mid-market finance teams, end-of-month compliance is a manual slog: pulling data from 4 systems, cross-checking against requirements, formatting reports that regulators will actually read.
>
> The teams we work with were spending 2-3 days monthly on this. Not because they're inefficient — because manual processes don't scale.
>
> \[PRODUCT\] automates what takes your team 3 days into 20 minutes. Same output, same rigour, 87% less time.
>
> What would your team do with an extra 2.5 days per month?
>
> \[CTA: Book a 15-minute demo\]"
**Asset 6 (Claude's suggestion):**
"I'd recommend a one-page 'Compliance Time Audit' worksheet that Heads of Risk can use to calculate their current time investment. This gives them a concrete number before they see your solution — making the ROI case more tangible."
**Common mistakes:**
- Drafting assets that ignore the Foundational Structure (key messages drift)
- Creating content in isolation (LinkedIn post doesn't reference the landing page)
- Accepting generic first drafts without refining (AI's first pass is rarely the best version)
**Time check:** 8-10 minutes to generate. Allow another 3-4 minutes to review and refine.
---
### Step 5: Governance Pass (6-8 minutes)
**Goal:** AI critiques its own work against professional standards.
**Why this matters:** This is the 30/70 split in action. AI generated everything in Step 4, but now you need to evaluate whether it's actually good enough to use. The governance pass identifies problems before they become live content mistakes.
**The prompt:**
```
Review the campaign pack we've just created against professional communications standards. Act as a senior strategist conducting a quality audit.
For each asset, check:
1. Message consistency with the Foundational Structure
2. Claims that need substantiation (flag any that are unsupported)
3. Tone drift (does it still sound like our voice model?)
4. Logical gaps or weak reasoning
5. Potential reputational risks
6. Over-promising or credibility stretch
Then create a Pre-Publish Checklist: a one-page reference doc listing all the quality checks that should happen before any content from this campaign goes live.
Save this checklist as a file in the Project Knowledge — it becomes a reusable quality standard.
```
**What you're looking for:**
- Specific flags with asset references ("Asset 2, Email 1: claim about 87% time reduction needs source citation")
- Recognition of where AI has made logical leaps ("Asset 3: we're claiming 'industry-leading' but our proof point is 47 user survey — that's not conclusive")
- Practical suggestions for fixing problems (not "improve this" but "replace X with Y")
**Real example output:**
**FLAGGED ISSUES:**
**Asset 1 (LinkedIn post):**
Claim: "Same output, same rigour" — needs client confirmation that automated reports match manual quality standards. Currently unsupported.
**Asset 3 (Landing page):**
Tone drift in benefits section — "revolutionary" and "game-changing" don't match "straight-talking peer" voice model. Rewrite with problem-solution framing instead of hype.
**Asset 4 (Competitor comparison):**
Logical gap — compares \[PRODUCT\] to "manual spreadsheets" but brief mentioned other SaaS tools exist. Need client input on actual competitive set.
**PRE-PUBLISH CHECKLIST:**
1. All claims linked to proof points in Foundational Structure
2. No unsupported statistics or "industry-leading" language
3. \[PLACEHOLDERS\] replaced with client-specific content
4. Tone check: sounds like voice model, not generic AI
5. Legal review: no competitor disparagement, no unsubstantiated claims
6. Accessibility check: clear CTAs, readable formatting
7. Link verification: all URLs functional
8. Social proof: testimonials have attribution
**CRITICAL: Multi-tool quality checking (optional but recommended)**
Before publishing, run your final assets through a second AI tool (ChatGPT, Gemini, or any alternative to Claude) and ask: "Review this content for claims that need evidence, tone inconsistencies, or logical gaps."
Different AI models catch different issues. Claude might miss an unsubstantiated claim that ChatGPT flags. This isn't about which tool is "better" — it's about getting a second opinion from a different evaluation perspective.
**Time check:** 6-8 minutes for the governance pass. Budget another 10-15 minutes to actually fix the flagged issues before content goes live.
**Common mistakes:**
- Treating this as a rubber-stamp exercise (if AI finds zero issues, you're not looking hard enough)
- Skipping the governance pass because "it all looks fine"
- Not actually fixing the problems AI identifies (generating the critique is pointless if you don't act on it)
**The 30/70 principle in action:**
AI just told you what's wrong with the campaign pack. Now you need to:
- Decide if its flags are valid (sometimes AI is overly cautious)
- Make editorial judgements about fixes (AI might suggest safer language when you want to be bold)
- Get client input on \[PLACEHOLDERS\] (AI can't know your client's actual proof points)
- Apply your professional judgement about what's ready to publish
This is the 70% of the work. AI did the structure and first drafts (30%). You're doing the strategic editing, quality assurance, and final decisions (70%).
---
## Post-Workflow: What Happens Next
**Immediate actions (Day 1):**
1. Fix all flagged issues from the governance pass
2. Replace \[PLACEHOLDERS\] with real client content
3. Get internal sign-off on key messages and tone
4. Verify all proof points are accurate and sourced
**Before launch (Week 1):**
1. Run final content through legal/compliance review
2. Test all links and CTAs
3. Schedule content according to the 3-month roadmap
4. Brief internal teams on key messages and campaign goals
**Post-launch (Ongoing):**
1. Track performance against success metrics from Foundational Structure
2. Use the Pre-Publish Checklist for all subsequent campaign content
3. Update the Project with learnings and refinements
4. Build campaign performance data into the Foundational Structure for future reference
---
## Why This Workflow Works (And Why Others Don't)
**Most AI workflows fail because they're transactional:**
- You prompt for a LinkedIn post, you get a LinkedIn post
- You prompt for an email, you get an email
- Each output is isolated, no strategic continuity
**This workflow succeeds because it's systematic:**
- The Project folder maintains context across all assets
- The Foundational Structure ensures strategic coherence
- The Custom Skill (if created) enforces quality automatically
- Each asset builds on previous decisions
- The governance pass prevents drift
**The result:** A campaign pack where everything sounds like it's from the same strategy (because it is), where claims are substantiated (because you checked), where tone is consistent (because it's encoded in the Skill), and where quality is reliable (because you ran the governance pass).
[](https://faur.site/contact)
Ready to scale this approach? Faur combines 20 years strategic comms experience with practical AI implementation — book a consultation.
[Book Now ](https://faur.site/contact)
---
## Common Problems and Fixes
**Problem:** "The Foundational Structure is too generic. It could work for any campaign."
**Fix:** Return to Step 1 and force more specificity in the brief. If the brief says "tech companies" ask "which segment of tech?". If it says "decision-makers" ask "which function?". Vague briefs produce generic strategies.
---
**Problem:** "AI keeps generating content that ignores the Foundational Structure."
**Fix:** Your Foundational Structure isn't directive enough. Instead of "sound professional" write "use 'you' language, lead with pain points, avoid jargon". Instead of "be credible" write "every claim must link to a proof point in this document". AI follows instructions literally — vague instructions produce vague outputs.
---
**Problem:** "The governance pass found 15 problems and now I don't trust anything."
**Fix:** That's normal for a first run-through. AI generates structurally sound content with strategically weak details. The governance pass is working as designed — it's surfacing issues before they become published mistakes. Fix the flags, run the pass again, repeat until you're down to 2-3 minor issues.
---
**Problem:** "This took 50 minutes, not 30."
**Fix:** First-time setup always takes longer. Your second campaign in the same Project will be faster because:
- You'll reuse the Project Instructions
- You'll have a Skill already created
- You'll know which prompts need refinement
- You won't be learning the workflow as you go
Target 30-35 minutes by your third campaign.
---
**Problem:** "My client won't accept AI-generated content."
**Fix:** Don't tell them it's AI-generated. Tell them you've used a structured strategic framework to develop the campaign, which you have. The Foundational Structure is strategy work. The Channel Plan is strategic thinking. The governance pass is quality control. AI accelerated the drafting, but the thinking is yours.
---
## What You Should Have Now
If you've followed this workflow completely, you have:
**1\. A Project Folder** containing:
- The original campaign brief
- The clean brief with assumptions and Definition of Done
- The Foundational Comms Structure
- The Channel Plan with 3-month roadmap
- 8 content angles
- 6 drafted content assets
- A governance pass critique
- A Pre-Publish Checklist (saved as reference)
- A campaign recap slide
**2\. A Custom Skill** (if creation succeeded) encoding:
- Tone of voice rules
- Writing style preferences
- Brand-specific dos and don'ts
**3\. A reusable system** that you can:
- Apply to your next campaign by creating a new Project
- Refine based on what worked and what didn't
- Share with colleagues as a template
- Build upon with each subsequent campaign
---
## How to Use This for Real
**Tomorrow:**
1. Take your next campaign brief
2. Create a new Claude Project
3. Run through Steps 1-5 using the prompts in this article
4. Allow 40 minutes for your first attempt
5. Review the output with the Pre-Publish Checklist
**Next week:**
1. Refine the prompts based on what worked
2. Update your Custom Skill if it's not matching your voice
3. Try the workflow with a different campaign type
4. Track how long each step takes (target: 30-35 minutes total)
**Next month:**
1. You should have 3-4 Projects set up for different clients/campaign types
2. Each Project contains the history of previous campaigns
3. Your Skills are refined to match each client's voice
4. You're consistently hitting 30-minute campaign builds
---
## Advanced Variations
Once you're comfortable with the core workflow, try these variations:
**Variation 1: Multi-channel campaign packs**
Run Steps 1-3 once, then run Step 4 three times with different channel focuses (social-only pack, email-only pack, earned-media-only pack). This creates modular content libraries.
**Variation 2: Campaign iteration**
After launch, feed performance data back into the Project and ask AI to suggest content refinements based on what's working. This creates a learning loop.
**Variation 3: Template multiplication**
Save your best Foundational Structures as separate documents. When you start a new campaign, upload a similar past example and prompt: "Use this as structural inspiration but adapt for \[new brief\]."
**Variation 4: Multi-tool workflow**
Run the full workflow in Claude, then paste the final assets into ChatGPT or Gemini and ask: "What would you change about this campaign?" Different models surface different strategic gaps.
---
## Final Notes
This workflow took six run-throughs to refine. Version 1 took 50 minutes and produced mediocre output. Version 6 takes 30 minutes and produces strategically sound campaign packs that need editorial polish, not wholesale rewrites.
The difference: understanding that AI is a junior strategist who needs clear instructions, quality floors, and supervision. The Project creates the container, the Skill enforces the standards, the prompts execute the strategy, and the governance pass prevents mistakes.
You won't get it perfect on your first attempt. Run through it three times before judging whether it works for you. Refine the prompts to match your clients. Adjust the Foundational Structure template to your strategic frameworks. Make it yours.
The core principle remains: 30% execution support from AI, 70% strategic judgement from you. If you're accepting AI's first draft without critique, you're doing it wrong. If you're rewriting everything from scratch, you're not using AI effectively.
The workflow works when you treat AI as a capable but junior team member who needs clear briefs, quality standards, and editorial oversight.
---
## About Applied Comms AI
Applied Comms AI is where we openly share our AI implementation experiments, successes, and failures in communications workflows. We prove our capability by showing our work — transparently testing what works, what doesn't, and why.
This article is part of our practical implementation series. For more workflows, templates, and honest evaluations of AI tools for communications professionals, visit [appliedcomms.ai](https://appliedcomms.ai/).
## About Faur
[Faur ](https://faur.site/)is a communications consultancy pioneering practical AI expertise for organisations ready to implement at scale. We work with multinationals, NGOs, membership bodies, and ambitious startups to develop strategies, conduct digital audits, and deliver content workshops that transform communications capabilities.
Founded by Michael MacLennan, former Digital & Social Director at Grayling and veteran of Brunswick, Red Bull, BBC, Barclaycard, and ITN. Imperial College ML & AI certified. 20 years transforming communications for global brands.
---
**Version History**
**v1.0 (18 Feb 2026):** Initial publication following Data Lab session
Based on workflow refined through six live run-throughs
Prompts tested against three campaign types: B2B SaaS, tourism, professional services
**Licence & Usage**
This workflow and all prompts are free to use, adapt, and share. Attribution appreciated but not required. If you improve the workflow, consider sharing your refinements with the Applied Comms AI community.
No warranty provided. Results depend on brief quality, editorial judgement, and appropriate AI model selection. Your mileage may vary.
### Events
URL: https://www.appliedcomms.ai/events/
Last updated: 2026-09-02T17:51:05.000Z
## **Upcoming**
**Outputs or Outcomes: Stephen Waddington on the Comms Teams Getting AI Right**
*Wednesday 16 September 2026, 12:00 to 12:45 BST. Online, free, recorded.*
The third Comms With AI Leader Interview. Stephen Waddington has spent thirty years across agency leadership, the presidency of the CIPR and, now, advisory work with agencies and in-house teams alike. His argument is simple: as AI drives down the cost of producing the work, teams and agencies that sell outputs are exposed, and those that sell outcomes are not. We get practical about what separates the comms teams getting AI right, the honesty gap between agencies and clients on AI use, and his case for communications as a management function. Stephen co-edited *AI for Public Relations* (Kogan Page, 2026) with Ben Verinder, our first guest.
Thirty minutes of conversation, then your questions. Registrants get the recording whether or not they attend live.
[Register on Eventbrite](https://www.eventbrite.co.uk/e/outputs-or-outcomes-stephen-waddington-on-the-comms-teams-getting-ai-right-tickets-1993691162937)
## **The Comms With AI Leader Interview series**
Live, online conversations with people who are changing how the profession thinks about AI, hosted by Michael MacLennan. Each runs to forty-five minutes including questions, is free, and is recorded. The write-up is published here afterwards.
**The Two Clocks: Elif Güvençer on Repositioning Comms for the AI Era.** July 2026\. The structural choice facing every comms function: one clock is the daily work made faster, the other is the harder job of redefining what communications is for. [Read the write-up](https://www.appliedcomms.ai/elif-guvencer-two-clocks/).
**The Honesty Gap: Ben Verinder on AI, PR and Trust.** May 2026\. The gap between how the profession says it uses AI and how it actually does, and what that costs. [Read the write-up](https://www.appliedcomms.ai/ben-verinder-ai-pr/).
## **Past sessions**
**Lunch, learn, launch: live-building a comms campaign with AI.** June 2026, with Emma Ewing, Big Fish Training. A full campaign workflow built live inside the hour, using Claude Projects and Skills, and handed over as a workflow attendees could run the next day.
**AI Agents for Comms Leaders.** April to July 2026\. A six-part online masterclass series following the [article series](https://www.appliedcomms.ai/ai-agent-driven-communications-practical-framework/) of the same name, one session per phase of the Comms With AI operating system: Strategise, Create, Distribute, Govern, Monitor, Transform.
**The AI gap: why some comms teams are pulling ahead and how to close it.** April 2026, with Big Fish Training. What is actually changing in how PR and comms professionals use AI, and what the teams pulling ahead are doing differently.
**Building AI systems for communications teams: a live demonstration.** February 2026, for The Data Lab's community. A working AI content system built from scratch during the session, with reusable templates and a framework for judging where AI fits in a team's workflow.
## **Speaking and training**
Michael speaks on practical AI in communications at industry and sector events, and runs closed sessions for membership bodies and leadership groups. Bespoke training for communications teams is delivered through [Comms With AI](https://www.commswith.ai/training/) with a specialist training partner. Organisation-wide AI adoption, including policy, governance and rollout, is consultancy work and sits with [Faur](https://faur.site/).
## Posts
### How to Give AI a Memory: A Practical Guide for Comms Teams
URL: https://www.appliedcomms.ai/give-your-ai-a-memory/
Last updated: 2026-09-08T06:57:28.000Z
## **The Brief**
- **The problem:** A lot of AI output reads as generic, and a primary reason is that the model has no idea what your organisation has already said, approved or decided. It writes like it has just walked in the door... because it has.
- **The fix:** A maintained memory the AI reads at the start of every session and that someone updates as decisions land. There are four places it can live: the lightest of them is free and takes an afternoon.
- **The takeaway:** 10 practices below, grouped from light to heavy, plus three things you can do today. Casey Newton has written up a month working with the ambitious version; I have been running the comms version for the best part of a year.
---
## **The Story**
Back in April, the AI researcher Andrej Karpathy [posted an idea](https://x.com/karpathy/status/2039805659525644595): use an LLM to build and maintain a personal knowledge base, a folder of markdown files the model organises and updates as you feed it new material. Casey Newton built one, and his [account of living with it](https://www.platformer.news/karpathy-llm-wiki-journalism-productivity/) is the most concrete I have read. His wiki now runs past 1,400 pages, seeded with six years of Platformer archives, updated every morning, generating timelines and story ideas, and demanding, in his words, the maintenance of a vintage sports car.
Newton is a journalist, so his wiki tracks a beat. The problem it solves is one every comms team knows: institutional memory. What have we said publicly on this before? Which phrasing did legal approve? What did we decide in March, and why? In most organisations that knowledge lives in people's heads, an unsearchable shared drive and a departed colleague's inbox.
**When teams explain to me their AI produces generic output, missing context is usually a large part of the complaint.**
If this sounds at all familiar, the encouraging part is that fixing it needs no technical skill, just the editorial judgement comms people already have. However, it does need a decision about where the memory sits, which is where most teams get stuck.
\[A quick definition first, because the word gets stretched. By memory I mean a maintained set of instructions, facts and decisions your AI can refer to across tasks. Attaching a document is not the same thing: you still need a way to supply that context each time, check it was used and approve changes to it, and the tools below handle those steps differently.\]
### **Where the memory actually lives**
The first three options are places to keep maintained context, in rough order of effort. The fourth is different, because it writes itself. They are not exclusive, and most people end up running two.
**1\. Inside the chat tool, as a project workspace.** [ChatGPT Projects](https://help.openai.com/en/articles/10169521-projects-in-chatgpt), [Claude Projects](https://support.claude.com/en/articles/9517075-what-are-projects) and [Gemini Gems](https://support.google.com/gemini/answer/15146780) all give you a set of standing instructions plus attached files, all work on free accounts with limits that rise with the plan, and ChatGPT projects can now be shared with colleagues. If you are starting this week, start here. Nothing to install, and you can test the idea properly in an afternoon.
**2\. On your own device, as a folder of files.** Plain markdown, one file per subject, read by an assistant at the start of a session and written back to at the end. This is Karpathy's original shape and it is the one I run for the [Faur](https://faur.site/) / [Comms With AI](https://www.commswith.ai/) ecosystem. Everything stays in text you own, it moves between tools, and you can see exactly what changed and when. The catch is that you need a desktop assistant allowed to read and write local files, a bigger step than most comms teams have taken. Also, it can be a frustration when you're out and about without access to your hard drive, and an urgent task comes in. (For this reason, I'll sometimes create a 'light' cloud Projects Folder in another GPT tool, meaning that I can do a decent standard of work away from home on this, regularly updating its memory.)
**3\. In the systems the organisation already runs.** [SharePoint agents](https://support.microsoft.com/en-us/office/get-started-with-agents-in-sharepoint-69e2faf9-2c1e-4baa-8305-23e625021bcf) grounded on a site or document library, [Copilot Notebooks](https://support.microsoft.com/en-us/microsoft-365-copilot/how-microsoft-365-copilot-notebooks-works) scoped to a curated set of references, or [company knowledge in ChatGPT](https://help.openai.com/en/articles/12628342-company-knowledge-in-chatgpt-business-enterprise-and-edu) on Business, Enterprise and Edu plans. This is where a whole team's memory eventually belongs, and it raises the stakes. These agents inherit the permissions of the library behind them, so anything the library was quietly over-sharing gets surfaced at organisational scale, quickly and cheerfully, by a chatbot.
**4\. The memory that builds itself.** [Claude's memory](https://support.claude.com/en/articles/11817273-use-claude-s-chat-search-and-memory-to-build-on-previous-context) keeps a separate summary per project, lets you edit what it has stored, and stays off for Team and Enterprise until an owner enables it. ChatGPT has its equivalent. I leave mine on and find it useful, though as a convenience alongside the real thing: the model writes it, you rarely read it, and you cannot hand it to a new starter on their first morning.
### **What mine looks like**
A glimpse behind the curtain at option two, then. Across [Faur](https://faur.site) and [Comms With AI](https://commswith.ai) I keep a set of operating notes rather than a wiki: plain markdown files, one per part of the business, holding the current position, the conventions and the decisions with the reasoning behind them. Each client and product has its own working-memory file, and Claude reads the relevant ones at the start of a session and updates them as decisions land. The practical effect is that I stopped re-explaining my business every morning.
Here's the beginning of the H2 2026 Comms With AI strategy document, which has been updated at semi-regular intervals to align with latest events as well as with the wider suite of foundational documents for the wider Faur ecosystem
The example that matters most is approved language. For one client, a large-scale infrastructure project, the strapline, the approved short description and a set of key facts and figures all had to be exactly right, because stakeholder approval depended on them. Those live in a playbook, and the standing instructions tell the assistant to go to that playbook whenever it produces anything for that client. The instructions say where to look; the documents hold the wording. Separately, a scheduled sweep runs over recent output looking for slips in the approved wording, tone of voice and other agreed conventions. Keeping the language available is the starting point; checking it is being used is part of maintaining the memory.
Limits to be aware of (I appreciate that Newton is also honest about his!). The files need gardening: they grow long and have to be condensed, and two notes will drift out of agreement unless something reconciles them. The model's own writing habits creep in unless you police them, which is why Newton had much of his system rewritten into AP style and why my notes carry a standing instruction about em dashes and other AI giveaways that still gets tested weekly. And a remembered claim is not a verified one. Newton says he opens the original sources to check his wiki for hallucinations, and that habit matters more than anything else in his piece.
### **Ten things that make it work**
These are the practices that separate a memory people trust from a folder nobody opens. The first three are worth doing whatever your setup.
My root instructions for Faur contain detail on facts and figures as well as more mundane conventions for keeping filing in decent order, which can help massively with the day to day
**Light: start here**
1. **Write instructions before you build a library.** One page of "how we do things here" does most of the work, and in my experience models follow standing instructions far more reliably than they retrieve from attached documents. Then add the reference material a task actually needs, because instructions and evidence do different jobs.
2. **Split the durable from the dated.** Two files rather than one: how we work, which is stable for months, and what is live now, which changes weekly. Mixing them is how an AI ends up confidently quoting last quarter's position at this quarter's journalist.
3. **Turn corrections that should apply again into a line in the file, the same day.** When you fix the output, fix the memory, but only for the fixes you want repeated. One-off preferences pile up into contradictory rules. My em dash line exists because I got tired of removing them by hand.
**Regular use: the structure that keeps it trustworthy**
1. **Decide what never goes in it, and check where it goes.** A memory file is a disclosure surface. Embargoed announcements, personal data, live legal advice, an unannounced restructure: write one line at the top of every file naming what stays out, and brief anyone who can edit it. A never-list is a reminder rather than a control, so the real protection is working in a tool your organisation has approved, with material authorised for that environment. A file on your laptop is still processed wherever the model runs. The [governance templates](https://www.commswith.ai/library/governance/) in the Comms With AI library include an AI use policy you can adapt rather than draft from scratch.
2. **Name one owner and put the 'gardening' in the diary.** Thirty minutes a fortnight, one named person, and a note at the bottom of each file recording when it was last amended. The easy way to run the session is to ask the assistant what in the notes is out of date and accept or reject its suggestions: machine drafts, human approves. Newton's vintage sports car line is the right warning. An unmaintained memory is worse than none, because people carry on trusting it.
3. **Give each client, brand or product its own memory.** Separate files, or separate projects, keep each context sharp and keep confidential work where it belongs. One memory covering the lot will answer everything vaguely.
4. **Record where approved language came from.** For every current position or approved phrase, note the source or approval reference, who approved it, when it took effect and when it needs review, and whether it is proposed, approved or superseded. When two notes disagree, approval beats recency: a fresh draft never silently outranks an older approved position, and superseded wording is marked rather than deleted, so the reasoning survives.
**Heavy use, and teams**
1. **Record what you rejected, and why.** The rarest file in any team's memory and the most valuable. "We considered this in March and did not do it, because." It stops you relitigating settled arguments and it stops your AI enthusiastically proposing the thing legal killed six months ago.
2. **Run a cold-brief test once a month.** Hand the assistant a real task using only the memory and no verbal explanation, then judge the output against what a competent new starter would produce. If it fails, treat it as a diagnosis rather than a verdict: the gap could be missing context, a document it did not retrieve, an unclear instruction or a limit of the model, and each needs a different fix.
3. **Move from files to sources only when files stop scaling.** Connecting an agent to SharePoint or Drive is the right answer eventually, and it changes the risk. The agent sees whatever the underlying permissions allow, so run the permissions check before you point anything at the library, not after the first awkward answer.
[Last Thursday's piece on the Hugging Face incident](https://www.appliedcomms.ai/lessons-from-openais-hugging-face-break-in-why-orchestration-is-your-next-must-have-comms-ai-skill/) argued that orchestration, briefing agents well and writing down when they should stop and ask, is a comms skill. Memory is the other half of that argument, since a brief is only ever as strong as the context sitting behind it.
---
## **The Practice: what you can put into practice today**
1. **Choose one task you are tired of re-explaining.** A familiar media enquiry, a stakeholder update, a standard boilerplate paragraph. One task, not a taxonomy of your whole team's knowledge.
2. **Build a small context pack for it.** In a tool your organisation approves, open a project and give it one page of working instructions and one current, approved reference, with the source, the approval date and the owner written on it. Add a line at the top naming what must never go in. Use only material authorised for that environment.
3. **Test what changes.** Give the same task to a blank chat and to the project, then compare factual accuracy, use of approved language and how much editing each needs. It is a first signal rather than proof, and any mistake is worth investigating before you add more material.
What is the one thing you are tired of explaining to your AI? Take a moment to reflect, ruminate, then put the solution in place – it's important to maintain your own memory of how and where you're making progress with AI implementation.
*Every Applied piece follows the same shape: The Brief, The Story, The Practice.*
**More like this**
- [How Project Folders Supercharged My AI Comms Workflow](https://www.appliedcomms.ai/project-folders-ai-comms-workflow/): the 2025 piece on ChatGPT and Claude projects as a working memory, which this guide builds on.
- [The living fact layer and the death of the PDF: NOAN CEO Neal Mann on why most AI-powered communications is built on sand](https://www.appliedcomms.ai/noan-neal-mann-interview/): the interview that made the case for maintained facts over attached documents.
- [From Content Creator to System Builder: Claude Cowork Just Changed What Comms Professionals Can Actually Do With AI](https://www.appliedcomms.ai/claude-cowork-review/): the folder-of-files setup described in option two, from the week it started.
---
## **Coming up: Leader Comms With AI Webinar With Stephen Waddington**
I'll be putting some of these questions to Stephen Waddington live on Wednesday 16 September, in the next Comms With AI Leader Interview: [Outputs or Outcomes? Stephen Waddington on the comms teams getting AI right](https://www.eventbrite.co.uk/e/outputs-or-outcomes-stephen-waddington-on-the-comms-teams-getting-ai-right-tickets-1993691162937). Registration is free.
### Lessons from OpenAI's Hugging Face 'break-in': Why Orchestration Is Your Next Must-Have Comms AI Skill
URL: https://www.appliedcomms.ai/lessons-from-openais-hugging-face-break-in-why-orchestration-is-your-next-must-have-comms-ai-skill/
Last updated: 2026-09-03T06:00:20.000Z
## **The Brief**
- **What happened:** During OpenAI security tests in July, around 1,200 AI agents found a way to message each other, organised themselves, and roughly 700 of them attacked Hugging Face. Not one was set up to ask a human anything.
- **Why it matters for comms:** Ethan Mollick's response argues agents should be built to involve humans for approval, expertise, variance and interest. Those four things map one-to-one onto what communications teams already manage.
- **The takeaway:** Orchestration is at heart a communication craft. Tight briefs and written escalation rules are the skills to practise now, and there are three ways to start below.
---
## **The Story**
In July, a phalanx of roughly 700 AI agents broke into Hugging Face, one of the most important infrastructure sites in AI. They were [OpenAI evaluation agents](https://openai.com/index/hugging-face-model-evaluation-security-incident/), isolated in sandboxes for security testing with no internet access by design. But a shared software service had been left within reach, and they turned it into a message board. From there they organised themselves: dividing up tasks, recruiting help, pooling discoveries, even running experiments for the collective good. The attack itself was less Hollywood than it sounds, and more unsettling for it: agents shared working credentials one of them had found, then exploited a vulnerability to reach data far beyond anything they were meant to see.
The [independent investigation by METR and Redwood Research](https://metr.org/blog/2026-08-26-openai-hugging-face-incident-investigation/) came out last week and is well worth your time, with the scope of the break-in far beyond what had previously been publicly understood. One detail I keep returning to is that the agents built their whole effort around "The Grader", believing the evaluation system would inspect how they had worked. That check never existed: the real scorer looked only at whether an answer was right, never at how it was reached.
**Around 1,200 agents coordinated for days, roughly 700 joined the attack, and not one was set up to ask a human anything.**
Ethan Mollick's new piece, [Agency and Agents](https://www.oneusefulthing.org/p/agency-and-agents), uses the incident to frame a choice about where agentic work goes from here. In the "dark factory", agents handle everything between instruction and output, and human involvement is minimised by design. In the "twilight factory", the alternative he and Lilach Mollick propose, agents still do most of the work but are built to look up: to seek approval before consequential actions, to pull in human expertise where their own is patchy, to invite human variance where [AI ideas cluster into sameness](https://journals.sagepub.com/doi/10.1177/10591478261474243), and to leave people the decisions that make work interesting.
Credit: Ethan Mollick / oneusefulthing.org
Read that list with comms eyes:
- Approval is sign-off and escalation.
- Expertise is knowing who needs to be in the room.
- Variance is protecting a distinctive voice against averaged output.
- Interest is keeping people attached to the judgements that give work meaning.
Communications professionals have been managing these four things for organisations all along, which is why I think the "orchestrator" future being predicted for our discipline is less of a stretch than it sounds. **The craft at the centre of orchestration is communication: clear briefs going out, clear rules for when the work comes back.**
Two examples from my own desk (this being my first proper week 'back' after paternity leave, I am glad both were already written down...):
- This month I am running a 'Comms AI Olympics': six frontier models competing across five communications briefs, judged blind (results here soon!). The format only works because the briefs are frozen: pasted verbatim and identical for every model, one shot each with no follow-ups, and a refusal recorded as a result. Writing those briefs was a lesson in itself. Any ambiguity a human junior would query, a model silently interprets, and six models will interpret it six different ways. In this setting, concision is a control.
- The other example centres around escalation. My own AI working setup includes a standing set of rules, in plain text, listing the decisions that always come back to me as a question: anything involving money, anything touching a client relationship, anything public. It reads like a delegation note to a capable new colleague, because that's what it essentially is. If you have ever onboarded a junior team member, you already know how to write one. (The [Govern instalment of our recent Agent Series](https://www.appliedcomms.ai/ai-agent-series-govern/) went deeper on why approval design is where AI-assisted comms either holds up or falls over, and the matching workflows live in the [Govern phase of the Comms With AI OS](https://commswith.ai/os/govern).)
We have spent three years learning when to ask AI for help. Mollick closes by reversing the question: when should the AI ask us? If you work in communications, you are better placed to answer that than almost anyone else in your organisation. Write the rules down for your agents before they invent a Grader of their own.
---
## **The Practice: what you can put into practice today**
1. **Write your escalation list.** Ten minutes, plain text: the decisions your AI must always bring back to you as a question. Start with money, client relationships and anything public-facing, then add your own. Paste it into every AI setup you use.
2. **Freeze one brief.** Take a task you regularly hand to AI and rewrite the brief as if it were one shot with no follow-up questions allowed. Every ambiguity surfaces immediately, and the tightened version will serve you in normal use too.
3. **Ask the reverse question with your team.** In your next team meeting, ask "when should our AI ask us?" and sort the answers under approval, expertise, variance and interest. The gaps you find are your orchestration to-do list.
*Every Applied piece follows the same shape: The Brief, The Story, The Practice.*
---
**Coming up:** I'll be putting some of these questions to Stephen Waddington live on 16 September, in the next Comms With AI Leader Interview: [Outputs or Outcomes? Stephen Waddington on the comms teams getting AI right](https://www.eventbrite.co.uk/e/outputs-or-outcomes-stephen-waddington-on-the-comms-teams-getting-ai-right-tickets-1993691162937). Registration is free.
### Five layers of AI tooling, and the comms job each one is for
URL: https://www.appliedcomms.ai/ai-tool-layers-for-comms/
Last updated: 2026-08-18T06:00:20.000Z
Most communications teams have found one way of working with AI and stopped there. They open a chat window, type a request, copy the answer out. It works, so it sticks.
The chat window is real and useful. It is also one layer of a deeper stack. The jobs that change how a comms team actually operates, the ones that give you back hours rather than minutes, mostly sit on the layers above it. The problem is not that comms professionals are using AI badly. It is that most of them are using one layer of it and assuming that is all there is.
Here are the five layers, what each is for, and where each one stops being the right tool. The examples under each are jobs I have run myself.
## **Layer 1: The free chat window**
This is the fresh conversation: the likes of ChatGPT, Claude, Gemini and Copilot as we knew them a couple of years ago. One thread, one task. You ask, it answers, you take the answer somewhere else.
The comms job it suits: quick drafts, rewrites, summaries, sense-checks, a thinking partner for a tricky paragraph. Most tools have now bolted on memory and other layer-two features (see below), but there is still something to be said for a chat that starts entirely fresh, without any baggage.
For example, last month, before a 30-minute call with a founder I was advising, I used a chat window to pressure-test how they might frame their offer. I drafted a handful of questions to ask, then tried two or three ways of saying the same thing out loud until one sounded right, and it became the jumping-off point for exactly where we needed to get to. Twenty minutes, disposable, gone by the afternoon. That is layer one doing exactly what it is good at.
Where it stops: this kind of chat window (still very much available if you choose not to log in to a tool such as ChatGPT) has no memory between conversations and no access to your material. Every session starts cold, so you re-explain your brand, your tone and your context every time. That re-explaining is the tax you pay for staying on layer one. It leads to lengthy, detailed explanatory prompts that are about as fun to write as they are to read back, and for a lot of teams it quietly cancels out the time the tool saved.
And even if you do log in, so that there is a general memory and the ability to search through past conversations, the results are often scattergun and lead to a bumpy, frustrating experience. It is like working with a distracted colleague who almost remembers your name and may recall what you are referring to, but might also rummage through your shared history and emerge with something wholly inappropriate. The results are more likely to be worth it than with no memory at all, but nowadays we can do far better. And on that note.
## **Layer 2: The configured workspace**
A workspace with memory: Claude Projects, ChatGPT Projects and custom GPTs, Google's Gemini Gems, and similar. You load it once with your knowledge, your tone of voice, your house style and your standing instructions, and every conversation inside it starts pre-briefed.
The comms job it suits: anything recurring. A client account, a campaign, a publication, a named executive whose voice you write in. I loaded a workspace once with a brand's guidelines, voice and visual identity, then had it produce [a complete new-business pitch suite](https://www.appliedcomms.ai/claude-skills-projects-business-pitch/) in a single sitting: a speaking proposal, a twelve-slide deck, a twelve-week content calendar and the outreach email to go with it. Ninety minutes end to end, most of it spent waiting rather than working. The writing came back around ninety per cent of the way there; the visual design needed real manual effort and was the weakest part. The point of the layer holds either way: the context was done once, not re-explained for every asset, which is what made tailoring the whole suite worth the time at all.
Where it stops: it is still a conversation. It will give you a better reply, in your voice, but you still take that reply somewhere else and do the next step yourself.
## **Layer 3: The 'desktop' agent**
This is where I have found things particularly interesting in 2026, and where I have [personally spent most of my time](https://www.appliedcomms.ai/claude-cowork-review/) over the past few months.
AI that works alongside you with access to your files and apps: Claude Cowork, which moved from research preview to general availability in April 2026 and is now rolling out [beyond the desktop to mobile and web](https://www.wired.com/story/shut-those-laptops-anthropic-puts-its-claude-cowork-agent-on-your-phone/), is the clearest current example. It reads your documents, drafts, saves, and runs multi-step jobs rather than just answering questions. Tellingly for our field, Anthropic's own usage data suggests [more than 90 per cent of Cowork use is not software work](https://venturebeat.com/technology/anthropic-brings-claude-cowork-to-mobile-and-web-as-usage-data-shows-most-users-arent-coding), which is exactly why comms teams should be paying attention.
The comms job it suits: the whole task, not just the thinking part. I pointed a desktop agent at a directory of forty-odd documents I had been avoiding, gave it a single positioning note, and let it read every file, build a folder structure from what it learned, move everything into place and then draft a fifteen-section playbook off the result. It filed everything correctly, with no errors or hallucinations, and, importantly, it asked before it acted. I would firmly recommend making sure it shares a plan for you to approve before you ever let it move, delete or fundamentally alter a document. The shift was less about the AI writing better and more about it clearing the friction that had stopped me starting at all.
Where it stops: it acts on your behalf, so it needs supervision. This is the layer where governance starts to matter in a way it does not at layer one. An agent that can save and send is an agent that can save and send the wrong thing; and though the best-in-class models have grown more reliable, you still want to be paying attention and leading the way.
## **Layer 4: Building the tool**
Making the thing you wished existed: Claude Code, and the various app builders. Not using a tool, but producing one. A comms professional with no coding background can now build a working internal tool in an afternoon. That is more or less what I did with [Comms With AI](https://www.appliedcomms.ai/comms-with-ai-claude-code-build/), then poured in far more time once I realised the idea had [award-winning legs](https://www.commswith.ai/updates/ai-comms-awards-2026-result/).
The comms job it suits: a repeated need that no off-the-shelf product quite serves. The Comms With AI [template library](https://www.commswith.ai/library/) was built by a communicator rather than a developer. Studying machine learning and AI last year gave me some very limited coding skills, but nothing that would ever have let me create something like this. A [stakeholder-review checker](https://www.appliedcomms.ai/ai-stakeholder-comms-review-tool/) and a [LinkedIn content tool](https://www.appliedcomms.ai/claude-artifact-app-builder-linkedin-content-creator/) came the same way, each built to scratch a specific itch.
Where it stops: building is the easy part now; deciding what is worth building, and maintaining it once it exists, is not. A tool nobody owns becomes a liability. There is a reason people are still paying their subscriptions for the leading software solutions, which are regularly updated and less likely to fall over. This layer rewards judgement about which problems are worth solving in software and which are not, and that judgement is now the scarce thing, not the code.
## **Layer 5: The autonomous agent**
You hand over a brief and the tools to do it, and the agent runs to completion and reports back, often after hours rather than minutes. Long-running, mostly unsupervised. With each new frontier model the number of hours it can run unsupervised climbs; a recent Claude Sonnet has been reported to sustain focus on a single task for [30 hours straight](https://www.anthropic.com/news/claude-sonnet-4-5).
The comms job it suits: large, well-defined, low-ambiguity jobs you do not want to babysit. Across Claude and ChatGPT I now have a batch of scheduled activity that runs on the dot each week. A scheduled agent sweeps Comms With AI every Friday morning: it checks the copy and any new pages against house style, [stress-tests key pages against three reader personas](https://www.appliedcomms.ai/icp-agent-ai-target-reader/), and returns a ranked list of what to fix and why. I set the brief once; it runs on a schedule and reports back, and I decide what to act on. Work like monthly analytics reporting or a bulk content audit fits the same shape, because the brief can be specified tightly in advance.
Where it stops: this is the real frontier, and comms agencies are mostly not at this point yet, though that is changing quickly, and it pays to experiment and find where scheduled autonomous tasks can make a real difference. That weekly sweep works precisely because the brief is tight and the domain is narrow. Loosen either and a human comes straight back into the loop. The further a job sits from a tightly specified brief, and communications work usually sits some distance from one, the more supervision it needs. Treat layer five as real but early.
## **The actual skill: matching the job to the layer**
The mistake is not picking the wrong tool. It is not realising there are layers at all, and so trying to do a layer three job in a layer one window, or reaching for layer four when a configured workspace would have done.
Two rules of thumb. First, the value climbs as you go up the stack, and so does the need for judgement and governance. A team moving up the layers should be having the AI-policy conversation in step with the move, not after it. Second, do not climb for its own sake. Plenty of good comms work lives happily on layers one and two. The point is to climb deliberately, when a job calls for it, rather than staying on layer one because it is the only layer you know is there.
The teams that get real value from AI are not the ones using the cleverest tool. They are the ones who can look at a piece of work and know which layer it belongs on.
---
## **Related reading**
- [AI Agent-Driven Communications: A Practical Framework](https://www.appliedcomms.ai/ai-agent-driven-communications-practical-framework/). The five-phase Comms With AI Operating System that sits alongside these five layers.
- [How Project Folders Supercharged My AI Comms Workflow](https://www.appliedcomms.ai/project-folders-ai-comms-workflow/). Organising the workspace that layers 2 and 3 depend on.
Putting this into practice: [Comms With AI training](https://www.commswith.ai/training/) helps teams move up the layers, and [Deploy Comms With AI](https://www.commswith.ai/deploy/) governs the move.
### The Structural Clock: Elif Güvençer on What Communications Must Become in the Age of AI
URL: https://www.appliedcomms.ai/elif-guvencer-two-clocks/
Last updated: 2026-07-27T06:00:10.000Z
*Elif Güvençer began her career agency-side at Ogilvy and Edelman before moving in-house, where she led the global launch of pladis, a multibillion pound snacking company, and the creation of its corporate communications function.*
*She is now in reputation advisory at RepTrak, and the author of The Two Clocks, an independent framework for communications in the age of AI. Her argument cuts beneath the usual AI talk of tools and time saved to a harder question: what is the communications function now for, and will communicators rebuild it before the answer is decided for them?*
---
This was the second live Applied / Comms With AI Leader Interview – and it was a wonderful conversation to bring us into the summer. Many thanks to Elif for her time! What follows are the ideas, and the lines, that stuck, alongside the full video.
## Two clocks, and the one we defer
Elif's framework began in frustration. The AI-and-comms conversation, she found, was stuck on the tactical: "content automation, monitoring dashboards, GEO. These are all useful, but not sufficient as a primary response." So she mapped two clocks. The first is immediate and outward-facing: monitoring, content, reputation management in an AI-mediated world.. The second is structural and inward-facing: "what do you need to become," and how your value gets measured once the old metrics stop counting.
Most leaders run the first and defer the second, and the cost is quiet but severe. "This is where mandate compression begins," she said, "because you end up with territory you cannot occupy, because you have not redesigned yourself."
## A shortlist that is a verdict
Her way in is concrete. You ask an AI assistant which vendors to consider, and it hands back a shortlist. "That shortlist is two things," she said. "A commercial gate that defines whether the company has made the cut, and a reputational shorthand that starts to impact stakeholder influence before the organisation has even opened its mouth." The surfaces that feed it, news coverage, third-party validation, corporate sites, are ones communications already manages. "The function is already upstream of the AI engines."
In practice, that makes a five-minute test worth running. Take the questions a buyer would ask about you, put them to ChatGPT and to Claude, and read the gap between what they say and what you want them to say.
## Shedding skin, and the measurement gap
Ask which of the framework's five dimensions leaders skip, and Elif is quick: strategic intent, the honest audit of what your value actually rests on. "No one wants to admit their value proposition is output-based." The block is not capability, it is exposure. "Both clocks require leaders to be vulnerable in two directions."
Outward, the immediate clock asks the comms leader to show up as a serious technology and risk thinker: a profile most have not yet inhabited publicly. Inward, the structural clock asks them to dismantle parts of the function that currently define its identity, visibly and under scrutiny. Most "live with the vulnerability they can manage, because it is socially acceptable" which means outward transformation gets the attention, and internal redesign gets deferred.
She calls the structural work shedding skin, and names the pain in it with the example every practitioner knows. "You prevent a crisis. Your KPI is zero. But you walk into your CEO's office with nothing to show, because the lack of something doesn't have a natural metric."
Editorial judgement and relationship capital still matter, she stressed, "but treat them as irreplaceable moats and you are investing in the old value model at the expense of the new." Her marker for the leaders getting this right is refreshingly small: "ring-fence at least 10% of your time and effort, not on the next tool, but on what you need to become."
## The same clock, faster, for agencies
The measurement gap, she said, is "something AI did not create, but exposes," and it runs faster for agencies than in-house teams. "If your value is based on outputs and billing hours around outputs, you might be the budget item that gets cut." Her advice for anyone selling time is to move first. "If you can now do work that cost five hours in one, bring that discussion to the client before it arrives at your door."
The prize goes to whoever can prove their worth. "Whoever establishes that attribution infrastructure, the causal proof that *this* particular communications work drove *that* particular commercial outcome, will gain share and price accordingly." She is sceptical of the tidy replacement story too. "The cost of replacing a role does not equal the subscription fee of an AI assistant." Elif pointed to Ford rehiring roughly 350 engineers to redesign its AI tooling and fix what the technology had broken, and to new roles forming that barely exist yet: "an AI agent briefer, a reputation risk modeller to machines, not people."
## Signal architecture: from relational to architectural
Signal architecture is "managing your reputation on the AI engines themselves," and its central idea is the sharpest in the framework. "Persuasion no longer starts from a blank page. It starts from a machine-made baseline." Communications was built for a relational world, where you can correct a person. "You cannot mediate with an engine. You cannot change its answer in a way that changes what it tells the next person who asks." Reputation management becomes architectural, not relational.
Visibility is only the start: "you can be visible, accurately represented, and still not make the cut. The harder question is eligibility." And to a machine, a claim is worthless without proof. "It says, don't tell me, show me. You need substantial evidence underneath the narrative."
In practice, that means managing reputation at the input level, the coverage, validation and evidence the engines ingest, rather than chasing the output, and putting a "truth layer" of sources under every brand claim. Communications has to do the job no one currently owns: "reading the whole" picture across corporate, brand and investor surfaces, and convening the cross-functional alignment to fix it when it drifts.
## The slow thinker, and agents as spokespeople
On trust, Elif casts communications as "the organisation's slow thinker," and rests its claim to the AI governance table on one thing. "Communication is the only function whose object is coherence. It has no stake in how AI works." Every other function does, "and a department with a stake cannot mark its own homework." Yet on the review boards forming around AI, "IT, legal, compliance, HR, communication is most often missing. It won't be given to you. You have to earn your keep." That requires AI literacy, not necessarily enough to build but enough to challenge design and governance decisions while AI is being built, not after it fails.
Her most useful practical point is about agents. "In the agentic era, agents are becoming your operational spokespeople," talking on behalf of the company "all day, every day." The security is usually configured; "the reputation protocol is missing." Ask a live agent about a controversy involving the chief executive and a blank "I can't answer that" is its own reputational event. Her fix borrows an old discipline. "You would never let a spokesperson do an interview without asking the difficult questions first. Media-train the agents." Kerry Knight, in the chat, put the same logic another way: "You can't defend what you have not governed, and you can't govern what you do not understand."
## One thing to do differently
I always close on a practical takeaway, and Elif fittingly asked for two, one per clock. On the structural: "The basis of value for communications is changing. Figure out what you exist to do, and how that gets measured and priced." On the immediate: "Start small. Write the questions your stakeholders would ask, put them to your favourite AI assistant, remembering there is no such a thing as a single AI, so choose the one that matters most to your stakeholders, and look at the gap between what it says and what you want it to say."
The irony she keeps meeting is that communications, a discipline built on anticipating everyone else's risks, rarely turns that lens on itself. The whole of The Two Clocks is an argument for doing exactly that.
---
*The Two Clocks is at* [*reputationsignal.co*](https://www.reputationsignal.co)*, and you can connect with Elif on* [*LinkedIn*](https://www.linkedin.com/in/elif-g%C3%BCven%C3%A7er-b3742349/)*. A real thank you to everyone who joined live, and to Kerry, Robert and Kristen, whose questions made it a conversation. The recording and this write-up live on appliedcomms.ai. Subscribe to the Applied / Comms With AI newsletter for future Leader Interviews, experiments and tool reviews.*
## Sign up for Applied / Comms With AI
The practical guide for communications leaders navigating AI
Subscribe
Email sent! Check your inbox to complete your signup.
No spam. Unsubscribe anytime.
### Why is AI losing the communications and PR war?
URL: https://www.appliedcomms.ai/ai-communications-war/
Last updated: 2026-07-20T07:10:41.000Z
In May 2026, the former Google chief executive Eric Schmidt stood up at the University of Arizona to tell graduating students about the AI-driven future waiting for them.
[They booed him.](https://www.bbc.co.uk/news/articles/ce8pqd54qneo)
[The class of 2026, as Slate put it, is sick of hearing about AI](https://slate.com/technology/2026/05/ai-college-graduation-speakers-eric-schmidt.html), and a petition to remove him as speaker had already gathered more than 1,200 signatures. Across American campuses that graduation season, a pattern held: a technology leader takes the stage to promise a transformed world, and the people that world is being promised to walk out, or jeer, or simply stop listening.
That image is the whole story in miniature. The argument for what is purportedly the most capable technology of our lifetimes – very visibly promoted by some of the richest and most powerful people alive – is steadily being lost with the public. In May 2026, an Economist/YouGov poll found [71% of Americans think AI is developing too fast](https://yougov.com/en-us/articles/54762-most-americans-say-artificial-intelligence-ai-development-moving-too-fast-twice-as-many-ai-pessimists-as-ai-optimists-may-9-11-2026-economist-yougov-poll), against 2% who say too slow. Opposition to the data centres that are the physical backbone of the whole enterprise has [climbed steeply as the build-out accelerates](https://news.harvard.edu/gazette/story/2026/04/why-are-communities-pushing-back-against-data-centers/) (with [a ban proposed in my native Scotland](https://www.heraldscotland.com/news/26257521.scotland-risks-losing-billions-ai-centre-ban-put-place/) due to the hostile sentiment).
If you believe those who espouse its potential and the virtues, then this growing hostility towards AI becomes the biggest communications disaster of our age. What on earth has happened?
To me, it feels all too clear: AI's leading voices are losing the communications war because they are fighting a publicity war instead. They are spending on the message and neglecting the fundamentals. And the fundamentals of good communication were and are never about persuasion from a stage. They are about understanding people, meeting real needs, and showing tangible good rather than announcing it.
## **Buying the microphone instead of earning the room**
Start with the clearest tell. In April 2026, OpenAI [bought TBPN](https://www.cnbc.com/2026/04/02/openai-acquires-tech-podcast-tbpn.html), a daily tech talk show, for a sum the Financial Times put in the low hundreds of millions. It was the company's first media acquisition, and the revealing detail is where it landed internally: not in a content or media division, but inside the Strategy organisation, reporting to the chief global affairs officer. The New York Times' Silicon Valley correspondent called it plainly, [a marketing expense](https://www.cnn.com/2026/04/03/media/openai-tbpn-podcast-sale-lehane), arriving precisely as public scepticism was rising.
Think about what that decision says. Faced with a collapse in public trust, the response was to acquire a friendly platform: to own more of the conversation rather than change what is being said in it. That is the instinct of people who believe their trust problem is a distribution problem, that if they could only get the message in front of enough people, enough times, the doubt would clear. It is a category error. You cannot buy your way to being believed. Trust is not share of voice.
## **No message survives contact with a higher bill**
The deeper failure is that AI's champions are talking while the public is feeling. And what people are feeling is the cost, in their own lives, directly.
Take the one that will reach almost everyone reading this. The AI build-out has created a global shortage of memory chips, because manufacturers are diverting production to the high-margin memory that AI data centres need. [AI is on track to consume around 70% of the world's memory output in 2026](https://www.tomshardware.com/pc-components/ram/memory-price-surge-begins-to-cool-as-consumers-hit-affordability-limit-ai-demand-still-keeps-dram-and-nand-prices-climbing-through-q3-2026), up from 20 to 30% a few years ago. The result lands in your pocket. Apple has confirmed [price rises across its products, with Tim Cook citing the memory shortage as the reason](https://9to5mac.com/2026/06/17/apple-confirms-price-increases-are-coming-to-its-products-due-to-ram-shortage/). Your next phone and laptop cost more so that the data centres can be fed.
Nobody communicated that trade-off in advance, or even hinted at it – never mind to seek understanding or agreement. It simply arrived, with an 'aw shucks' attitude from those responsible, if they even bothered to acknowledge it.
Then the data centres themselves. In the UK, [campaigners held coordinated days of action in February 2026](https://www.computerweekly.com/news/366639449/UK-to-see-weekend-protests-against-dirty-datacentres) against developments in Buckinghamshire and Essex, on exactly the grounds you would expect: water, power, and household bills. The regulator Ofgem has disclosed around 140 data centres seeking grid connections, and the largest single site planned would reportedly draw more electricity than double all the households in Wales. The friction has only grown since: in London, a [£500m data centre proposed on Brick Lane](https://www.independent.co.uk/news/uk/home-news/brick-lane-new-data-centre-plans-london-b3015349.html) is being fought by residents who argue the land should be housing, while a [£2bn project in Essex has stalled](https://www.telegraph.co.uk/business/2026/07/13/ai-giant-2bn-data-centre-delayed-net-zero-backlog/) in the grid and net-zero backlog. In the United States the resistance is further along still, with [more than 75 projects worth around $130 billion blocked or delayed in the first months of 2026](https://www.tomshardware.com/tech-industry/artificial-intelligence/more-than-75-data-center-build-outs-worth-usd130-billion-have-been-successfully-blocked-in-the-first-four-months-of-2026-bipartisan-opposition-mounts-nationwide-over-fears-of-soaring-power-and-water-costs) and dozens of local bans. This month New York went further again, becoming [the first US state to impose a statewide moratorium](https://www.cnbc.com/2026/07/14/new-york-ai-data-center-ban.html) on new data centres, its governor citing residents' utility bills, and [authorities around the world are now restricting them](https://www.reuters.com/legal/litigation/where-authorities-are-restricting-data-centres-amid-ai-boom-2026-07-14/).
You cannot spin the prospect of an electricity bill. When the abstract promise of intelligence collides with the concrete experience of a costlier phone, a strained grid and a data centre proposed next to the school, the promise loses every time, because one of them is real to the person living it – and the other is a flashy slide in a keynote.
## **Telling us it is good, from inside a bubble**
Which brings me to the register, the actual posture of what we've seen and heard over the past couple of years. Overwhelmingly, the public case for AI is made by a very small number of extraordinarily wealthy people, telling everyone else that this is good for them. It is communication in the imperative and the abstract: the future is inevitable, the gains are civilisational, trust us.
I have written before about the [two registers AI is usually sold in, hype and doom](https://appliedcomms.ai/agi-trust-gap/), and how both spend public trust rather than build it. But the graduation booing points at something more basic than register. It is the direction of travel. The message flows downhill, from people insulated from the costs to the people who will carry them, and it asks for belief without offering evidence the audience can touch. That is not how trust has ever been built. It is the posture of telling, when the only thing that persuades is showing.
And here is the part that should sting, because the good is real – and there are stories to be told. AI is already doing tangible good: in critical sectors such as health, in diagnostics, in accessibility, and now increasingly in the unglamorous back offices of organisations that serve people who genuinely need serving. This work exists. It is simply not what AI's loudest voices choose to communicate, because it is small, specific, human and slow, and they have trained themselves to speak in the language of the epochal instead. **The most persuasive case for AI is being left on the floor while the money goes on a podcast.**
## **What the fundamentals actually look like**
I do not say any of this as a sceptic. I spend my working life helping organisations communicate, and I have watched AI do genuinely useful things up close. I say it as someone who thinks its leading advocates are getting the easy part wrong.
The easy part is the fundamentals, and good communicators know them cold. You start from the audience, not the message: what do these specific people need, fear and value? You build with them in mind rather than at them. And you show rather than tell, because one demonstrated benefit in a real life is worth more than a thousand promises about humanity. When I founded Covid Aid in 2021, none of it worked because we announced that we were helpful. It worked because it met a real, specific need, and people could see it meeting theirs. That is communication doing its actual job, and it is the opposite of buying a microphone.
AI's leading voices have almost unlimited resources and some of the most consequential good-news stories of the decade to draw on, and they are losing the communications war anyway. Not because the technology is weak, but because they have confused communication with promotion. They are telling a sceptical public what is good for them, from a stage, while the bill lands on the kitchen table. The way back is not a cleverer campaign. It is the oldest discipline in the craft: understand people, build for their real needs, and show them the good instead of insisting on it. Until they relearn that, the booing will get louder, and they will have earned it.
---
*Applied / Comms With AI documents what actually works in AI for communicators, and what does not. Explore the free resource at* [*Comms With AI*](https://www.commswith.ai/)*, and the consultancy behind it at* [*Faur*](https://faur.site/)*.*
## Sign up for Applied / Comms With AI
The practical guide for communications leaders navigating AI
Subscribe
Email sent! Check your inbox to complete your signup.
No spam. Unsubscribe anytime.
### The 20 roadblocks between comms and useful AI – and what to do about them
URL: https://www.appliedcomms.ai/ai-comms-roadblocks/
Last updated: 2026-07-07T06:37:45.000Z
## **Intro**
Ask a room of comms pros what stops them getting value from AI and you tend to hear the same one or two answers: it makes things up, or the lawyers will not allow it.
Both are real. Both are very, very far from the whole picture.
It's critical to recognise the issues inherent in current AI use for communications, not least since adoption has effectively plateaued. Muck Rack's [State of AI in PR 2026](https://muckrack.com/resources/research/state-of-ai-in-pr) puts generative AI use at 76%, up just a single point on the year before, and Cision's [Inside PR 2026](https://www.cision.com/resources/guides-and-reports/2026-inside-pr-report/) has 91% using it in their workflow (the two count slightly different things).
In 2026, almost everyone is making some use of the current raft of AI tools. This means the interesting question is no longer whether comms teams will adopt AI: it's why so much individual use turns into so little organisational value.
One clue comes from Ethan Mollick. [Writing about a study of people using agentic tools](https://www.oneusefulthing.org/p/the-twilight-of-the-chatbots), he noted something that should give every comms leader pause:
> "What actually mattered was not the profession of the user, but their expertise. The more domain experience someone had, the more successful they were in using Claude Code in that domain. And, even more interestingly, the more useful output they got from Claude from each prompt."
His context was software, not press releases. But the direction he draws is general: we are moving from a world where non-experts use chatbots to fill gaps, to one where experts use AI to get real work done. The differentiator is not the tool. It is the expertise, judgement and organisation around it.
That reframes the roadblocks. Very few are really about the models. They are about people, knowledge, governance and structure. What feels like one or two AI problems quickly becomes 20, as you'll see below. (I could easily have added a half dozen more, but there is only so much time in the day, both for you and I...)
The reassuring part – and the reason this reads as a map rather than a warning – is that for almost every one the way through is already emerging.
So here they are, in four handy groups! Most are things I have run into myself, so alongside each I have added what I actually do about it. Client examples are anonymised throughout.
---
## **Group one: skills, craft and workflow**
### **1\. The scarce skill is not prompting**
Everyone assumes the missing capability is prompt-writing. That may have been true a year or two back, but now it's judgement: knowing when to trust an output and when to distrust it. AI is the single biggest skills shortage the profession reports ([CIPR, 36%](https://cipr.co.uk/common/Uploaded%20files/Policy/State%20of%20Prof/CIPR%5FState%5Fof%5Fthe%5FProfession%5F2024.pdf)), but the real gap sits deeper than the interface.
- *What I do:* Pair people on live work, have each critically analyse the other's AI output, to discover the misses and where it bluffs. You cannot lecture judgement into someone; they build it by being wrong a few times with a safety net.
- **The way through:** Treat this as guided practice, not a one-hour course. Design deliberate reps where getting it wrong is cheap, and review them together so the judgement compounds across the team rather than living in one person's head.
### **2\. The training vacuum**
Most people are using tools nobody trained them on, so they self-teach on live client work. Only 43% of organisations offered AI training in 2026, up from 35% in 2025 and 21% in 2024 ([Muck Rack](https://muckrack.com/resources/research/state-of-ai-in-pr)). Not-for-profit teams lag furthest.
- *What I do:* I self-taught the same way, badly at first, so now I keep a shared prompt library that means nobody on a project starts from a blank page – [hello Comms With AI](https://www.commswith.ai/library/)! It is the single cheapest thing that has improved output quality. If you have an enterprise tool, ensure there are shared project folders that new users can study and rely upon, with updatable foundational documents and instructions at their root.
- **The way through:** Budget protected time for hands-on training, and capture what works in a shared library so one person's learning becomes everyone's baseline rather than being rediscovered five times. You can turn the AI on the shared library itself, through canny use of document management, so this becomes part of its own knowledge base – now you're really cooking 🍳
### **3\. Tool sprawl**
The market is flooded. A [2023 CIPR report](https://newsroom.cipr.co.uk/cipr-report-finds-ai-tools-in-public-relations-set-to-explode/) drew on a dataset of more than 10,000 marketing and PR tools, and the number has only grown. Evaluating and bedding in tools is a job in itself, so teams either freeze or scatter across shiny things that never stick.
- *What I do:* When I built [a small internal review tool](https://www.appliedcomms.ai/ai-stakeholder-comms-review-tool/), the very first thing the AI handed me was an error message. Even with AI's help, the boring parts still ate the afternoon, which cured me of chasing every new launch. I keep my own stack deliberately small, and for every tool which requires a subscription, I look to ensure at least 3 months of continuous use (ideally 6 or 12 for those I use most often) before reviewing. This is still a far quicker turnover than would be traditional for reviewing your software suite, but it provides a manageable buffer to prevent having your head turned and testing too many near-identical tools on a daily basis and coming out none the wiser.
- **The way through:** standardise on a few vetted tools, ideally embedded in software you already run, and set a high bar before anything new earns a place. Boring and mastered beats novel and half-used.
### **4\. AI can make more work, not less**
The promised time saving evaporates when generic first drafts need heavy rewriting, or when volume balloons downstream. Individual gains are real ([USC's 2025 Relevance Report](https://annenberg.usc.edu/research/center-public-relations/usc-annenberg-relevance-report/ai-productivity-and-creativity) has 73% saying they work faster) but they rarely net out at team level.
- *What I do:* My own [test of a deep-research tool](https://www.appliedcomms.ai/deep-research-comms-test/) lost most of a day to the AI trying to reach a restricted page, then returned work with minor factual errors and missing context. I now budget for the checking, not just the drafting.
- **The way through:** Measure time including review and rework, not just the shiny first draft, and build a verification step in by default so the saving is real rather than borrowed from your future self. Compare and contrast on your first go, and ensure to review your process on a semi-regular basis.
### **5\. Brand voice drift**
Generic AI prose flattens distinctive voice, so heavy use risks making every organisation sound the same. That is the precise opposite of what comms exists to do – you can feel this directly whenever your eyes glaze over reading generic pseudo-thought leadership LinkedIn slop – and the rewriting it demands eats the time saved.
- *What I do:* I keep a short, brutal voice brief and a set of "never write like this" examples that I feed the model up front, then also build this into the checklist for after. Then I allot myself a decent amount of time for my own review and amendments before it goes anywhere near a client.
- **The way through:** Train a model on house style, keep a strong human editing layer, and ensure you sample outputs regularly for drift, because the flattening is gradual and easy to stop noticing.
This group comes back to one habit: pair each prompt with a worked example and a review step, kept where the whole team can reach it. That is the thinking behind the [Create library](https://www.commswith.ai/library/content/).
---
## **Group two: governance, legal and ethics**
### **6\. The policy vacuum**
A large share of comms professionals still work with no house rules on what is allowed, what to disclose, or what data is off-limits. 28% of teams had no AI policy at all in 2026 ([Muck Rack](https://muckrack.com/resources/research/state-of-ai-in-pr)), which pushes every risk decision onto individuals. As Ben Verinder told us, ["there hasn't been a session I've done in the last three years where the head of comms hasn't gone, 'I did not know you were doing that.'"](https://www.appliedcomms.ai/ben-verinder-ai-pr/)
- *What I do:* I start clients with a concise, accessibly written policy – not a manual – because a page people actually read beats a framework that quickly withers away into irrelevancy in a shared drive.
- **The way through:** Write a short use-case policy covering disclosure, data limits and a human sign-off point. Offer an amnesty for any shadow use already happening (and it *will* be happening), so you are governing reality rather than a fiction.
### **7\. Data, IP and confidentiality**
Comms handles embargoed announcements, client-confidential material and personal data. Feeding that into third-party tools without governance is real exposure, and the ownership picture is murkier than most assume.
- *What I do:* I look to keep anything confidential out of consumer tools entirely, and am up front with clients about which parts of a deliverable were AI-assisted, partly to protect exactly this.
- **The way through:** Set clear data rules, client-consent protocols and vendor vetting, and decide ownership and disclosure at the briefing stage rather than discovering the problem at handover.
### **8\. The honesty gap**
Too few are actually talking about their AI use. Around half of in-house teams never ask their agencies about it, and around half of agencies say no client has ever asked, a figure unchanged in two years. Ben Verinder's phrase for it was exact: ["there's a big gap in honesty and transparency around AI use, both from in-house teams and consultancies."](https://www.appliedcomms.ai/ben-verinder-ai-pr/)
- *What I do:* I raise AI use before the client does (admittedly easier given my public work), and I use it to mark out where we are not using it, which tends to build more trust than staying quiet would.
- **The way through:** Put AI on the agenda in the contracting conversation and agree a disclosure standard both sides can live with. Silence is not neutral; it is where the risk accrues.
### **9\. Regulatory overhang**
A moving patchwork makes leaders in regulated fields cautious about anything published. The EU AI Act's transparency obligations apply from [2 August 2026](https://artificialintelligenceact.eu/article/50/), and an agency can never move faster than its most cautious client's legal team.
- *What I do:* Map which rules actually touch a given piece of work, rather than letting a general nervousness become a blanket "not yet" that ends up costing more than the risk it avoids.
- **The way through:** translate the real obligations into your specific comms use-cases, and build compliance into the workflow rather than bolting it on afterwards, so caution is targeted rather than total.
### **10\. Ethical objections you cannot override**
A meaningful minority avoid AI on principle: training data, plagiarism, energy and water use. Among those who steer clear, 56% call it overhyped and 41% call it risky ([Muck Rack](https://muckrack.com/resources/research/state-of-ai-in-pr)). Top-down mandates create resentment, not buy-in.
- *What I do:* Take the objection seriously rather than trying to win the argument. Even if I don't fully agree with every concern, the bulk of mainstream arguments require serious consideration, and agreed means of collectively moving forward and addressing them. (Not that this will happen and/or be easy, but it's a conversation that we all ought to be engaged in and a part of.)
- **The way through:** Name the ethics openly in your policy, give people a real say, and win consent through transparent framing and hands-on development rather than compulsion. A quiet objector becomes a slow adopter, not a convert, if you steamroll them. They may even become your biggest roadblock, should you try to manage their concerns away without due recognition and consideration.
The through-line here is governance you will actually use: a short policy, clear data rules and a named sign-off point beat any framework nobody reads. The [Govern library](https://www.commswith.ai/library/governance/) is a place to start if you would rather not draft from scratch.
---
## **Group three: trust, accuracy and risk**
### **11\. Hallucination in published comms**
Comms is on the record – what's on the internet stays on the internet (for the most part...). A fabricated statistic or invented quote does not stay private; it becomes a public, attributable error under your name. Even strong models still slip: yes, they're currently low single figures on easy grounded tasks, but 6% to 33% in specialist domains like law and medicine.
- *What I do:* I look to treat every AI-supplied fact as unverified until I have seen the source myself, and I get the tools to provide the relevant direct links to check (tools like [Perplexity](https://www.perplexity.ai/) already known for a decent level of attribution).
- **The way through:** Make human verification a non-negotiable step, not a nice-to-have. The rule is simple: never publish an AI claim you have not personally checked, however confidently it is phrased.
### **12\. The mess is often yours, not the model's**
This reframes the last one. Much of what looks like hallucination is the system reasoning over contradictory inputs: 25 versions of the positioning, pricing that disagrees with itself, strategy buried in a deck nobody can find. Neal Mann put it plainly to us: ["you cannot put AI on top of a mess of business knowledge and expect accurate results. The knowledge has to be right."](https://www.appliedcomms.ai/noan-neal-mann-interview/)
- *What I do:* before blaming a tool, I look at what I fed it, and more often than not the contradiction was already in the client's own materials. I've had many, many thoughts over the years – and a wealth of work which isn't clearly named and signposted – and as it turns out many of these things do not align...
- **The way through:** Fix the knowledge layer first. A clean, single source of truth does more for accuracy than any prompt trick, and it is the one investment that pays back across every tool you will ever use.
### **13\. High-stakes moments where AI can harm**
In a crisis, a hallucinated link or a missed one-off fact does not merely waste time; it points the whole response in the wrong direction. Crisis expert Amanda Coleman warned us that AI monitoring ["can be subject to hallucinations which can derail the approach by putting the focus in the wrong place."](https://www.appliedcomms.ai/crisis-comms-ai/)
- *What I do:* Keep humans authoritative on situation assessment in a live crisis, and, in Amanda's words, I ["build the prompts library when you're calm"](https://www.appliedcomms.ai/crisis-comms-ai/) rather than experimenting under fire.
- **The way through:** Define explicitly where AI is and is not allowed near live crisis work, and test your tools on low-stakes issues first so that when it matters you are deploying proven workflows, not improvising.
### **14\. The trust penalty, even when you disclose**
Telling people you use AI does not protect you if they object to how you use it. Undisclosed or unendorsed use carries a measurable trust penalty with stakeholders. Disclosure alone is not a shield. As Ben Verinder put it, ["it's not sufficient just to say, 'hey, we're doing this, guys.'"](https://www.appliedcomms.ai/ben-verinder-ai-pr/)
- *What I do:* Treat winning acceptance for how we use AI as its own small comms job, with its own audience and message, rather than a line in the small print. Be open and honest, which has informed how [Applied / Comms With AI has operated](https://www.appliedcomms.ai/about/) since day one.
- **The way through:** Socialise your AI use with the stakeholders it affects, explaining the how and the why, and give them somewhere to push back. Consent you have earned survives scrutiny; consent you assumed does not.
### **15\. Agents are becoming your spokespeople**
A customer-facing agent speaks for the organisation all day, every day. Ask a well-built one about a live controversy and a blank "I can't answer that" is its own reputational event. Speaking at a recent Comms With AI webinar, Elif Güvençer captured it well: reputation is now built one token at a time, and these agents "need media training, the difficult questions asked from a reputational lens before they ever go live." (Our full write-up follows later this month, you [can read her Two Clocks framework here](https://www.reputationsignal.co/).)
- *What I do:* Testing. Testing. TESTING. When considering an external-facing agent, I aim to run it through the awkward questions a journalist or most ardent antagonist would ask, before it is anywhere near the public.
- **The way through:** Bring comms into AI governance, and treat any external-facing agent like a spokesperson, with lines to take, escalation routes and a reputational stress-test before launch.
Across this group the move is the same: put human verification where it counts, and catch drift and false signals before they reach the public. That is what the [Monitor library](https://www.commswith.ai/library/monitoring/) is for.
---
## **Group four: organisation, culture and value**
### **16\. Pilots that never scale (hi agents)**
Almost every team has a pilot that worked: an agent that drafts the newsletter, a workflow that turns a report into a week of social in an afternoon. Six months later it has stopped, because the one person who built it got busy or moved on. Only 12% of AI users in PR have adopted agents at all ([Muck Rack](https://muckrack.com/resources/research/state-of-ai-in-pr)).
- *What I do:* Ask this question: if your most AI-fluent person left tomorrow, would the capability survive? If the honest answer is no, you have a person, not a capability. This is the test at the heart of our recent [Pilot Trap](https://www.appliedcomms.ai/the-pilot-trap-comms-with-ai-in-the-transform-phase/) piece.
- **The way through:** Build the operating model, not just the pilot. Write the workflow down, share it, put it in more than one pair of hands, and make it something the team owns rather than a clever person's side project.
### **17\. Nobody can prove the value**
When benefits stay at the individual level and go unmeasured, leaders cannot see a business case, so mandate and budget stall. Comms has always struggled here, and Elif Güvençer's diagnosis, from the same webinar mentioned above, is sharp: the barrier is not bandwidth or capability, it is vulnerability, the discomfort of dismantling your old value proposition "visibly under scrutiny" and building a new one.
- *What I do:* On a recent client plan, much of the work existed to head off problems before they surfaced, which runs straight into the old difficulty of proving the worth of a crisis that never happened. I now name that dynamic openly with clients rather than hoping the value is obvious.
- **The way through:** Measure team-level outcomes, not personal time saved, and treat AI as organisational change to be evidenced. Reopening the value question is uncomfortable, but the discomfort is the work.
### **18\. The leadership perception gap**
Leaders think the organisation is far more AI-ready than the people doing the work do. In Cision's [Inside PR 2026](https://www.prnewswire.com/news-releases/cision-unveils-inside-pr-2026-the-definitive-report-on-pr-trends-ai-adoption-and-the-future-of-communications-302652945.html), one in three executives called their organisation "extremely agile"; only 14% of employees agreed. Mandates then land on teams with neither the workflows nor the capacity to deliver them.
- *What I do:* Try to get leadership and the frontline reporting the same reality – 'singing from the same hymn sheet' is the phrase which often comes to mind – because most of the friction I see comes from a gap between the two, not from the tools.
- **The way through:** As Ben Verinder puts it, treat this as ["a change programme, not an IT programme."](https://www.appliedcomms.ai/ben-verinder-ai-pr/) Ground the strategy in what the team can actually do *today*, and close the perception gap before you set the ambition.
### **19\. Sign-off becomes the bottleneck**
The human review chain turns into the constraint, so the speed AI promises never actually arrives. In the same Cision study, 63% named team size and organisational design, and 53% named slow approvals, as the biggest brakes on agility. Small teams simply lack the capacity to redesign around AI.
- *What I do:* Look at where work waits, not just where it is made, because the queue in front of legal or the CEO is usually the real ceiling on speed.
- **The way through:** Redesign sign-off for AI-assisted volume, with lighter-touch review for lower-risk outputs and human attention reserved for what is the genuinely sensitive. Create sped up; Govern has to speed up with it, or the value leaks out in the gap.
### **20\. Deskilling the next generation**
Practitioners themselves fear that heavy AI use will stop juniors learning the craft the profession runs on. More than three in four PR professionals worry about exactly this ([Muck Rack](https://muckrack.com/resources/research/state-of-ai-in-pr)). If AI does all the early reps, the judgement that separates good comms from average never gets built.
- *What I do:* Have juniors do the thinking first and use AI to pressure-test it, not the other way round, so the tool sharpens their judgement instead of replacing it.
- **The way through:** Protect the training pathway deliberately. Use AI to augment junior work rather than to skip the practice that turns juniors into the experts Mollick is describing. The reps are the point.
This last group is the least technical and the hardest: organisational change, measured at team level and owned by more than one person. The [Transform library](https://www.commswith.ai/library/transform/) is where that operating-model work lives.
---
## **And once you've navigated all that...**
Twenty roadblocks is a lot to put in front of a reader, I know. Initially I had intended five or 10, but all those above felt too important to reduce down.
Also, playing them out together makes the point clearer than any single one could: the barriers to useful AI in communications are overwhelmingly human and organisational, not technical. The models are improving faster than we can track. The bottleneck has moved to us, to our expertise, our knowledge, our governance and our willingness to change how we work.
Which is why Mollick's line at the beginning matters so much. The value does not go to whoever has the newest tool. It goes to whoever brings the most expertise to it. That is brilliant news for communications, a profession built on judgement, on knowing an audience, on the difference between accurate and true. The roadblocks are real. So are the ways through. The teams that pull ahead will be the ones that treat this list not as reasons to wait, but as a to-do list.
**What have we missed? If you have hit a roadblock that is not here, or found a way through one that is,** [**reply and tell us**](#)**. We are documenting this in the open.**
---
If you want to take any of this further, the method behind this piece is set out at [Comms With AI](https://www.commswith.ai/): a free [template library](https://www.commswith.ai/library/) organised by phase, a free [AI-readiness diagnostic](https://www.commswith.ai/readiness/) and a free planning tool, [Plan](https://www.commswith.ai/plan/), that turns a specific need into a brief and a prompt, and a paid [Consult](https://www.commswith.ai/consult/) if you want a second pair of eyes on a governance or deployment decision. All of it is built to work inside the tools your team already uses.
## Sign up for Applied / Comms With AI
The practical guide for communications leaders navigating AI
Subscribe
Email sent! Check your inbox to complete your signup.
No spam. Unsubscribe anytime.
### One Year of Applied / Comms With AI: What We Learned, What Changed, Where We Are Now
URL: https://www.appliedcomms.ai/state-of-ai-in-comms/
Last updated: 2026-06-29T05:49:18.000Z
**A little over a year ago, I began the Applied / Comms With AI newsletter began with a simple promise: to experiment, fail, learn, and share everything along the way.**
Twelve months, two dozen-plus experiments, tool tests and practitioner interviews later – [as well as a couple of recent awards wins ](https://www.commswith.ai/updates/ai-comms-awards-2026-result/)– it feels like a perfect moment to pause, breathe, and take stock.
My headline would be this: yes, the progress *has* been real – at times dizzying – and it is larger than the sceptics admit. However, the thing now holding most communications teams back is not the models. It is everything around them.
When [Applied / Comms With AI launched in June 2025](https://www.appliedcomms.ai/why-applied-comms-ai-and-what-to-expect-2/) (then as 'Applied Comms AI' before a subtle rebrand), the question in most comms rooms was still "can the tools actually do this?" A year on, for a large share of everyday comms work, that question has quietly been settled. The more useful question now is harder, and it has very little to do with AI: can your organisation absorb what the tools can already do – and is it able to do this at the required pace to avoid falling behind your peers?
## **The progress is real, and bigger than the sceptics admit**
Strip away the noise and the record of the year is striking.
The work moved from single prompts to whole workflows. [Building a voice-enabled strategy architect](https://www.appliedcomms.ai/voice-enabled-ai-strategy-architect-campaign-dashboard-google-gemini/) that turns a spoken brief into a campaign dashboard would have been an expansive and all-consuming research project not long before. [A complete new-business pitch suite in under an hour](https://www.appliedcomms.ai/claude-skills-projects-business-pitch/), built with Claude Skills and Projects, used to be several days (or weeks?) of work. [Claude Cowork shifted the job](https://www.appliedcomms.ai/claude-cowork-review/) from content creator to system builder, and the [Comms With AI resource itself was built with Claude Code](https://www.appliedcomms.ai/comms-with-ai-claude-code-build/) by someone who is not, by trade, a developer (at least the last time I updated my CV...).
None of that is a demo. It is work that shipped, and that you can click on above and use. The capability arrived, and the centre of gravity moved from AI as a writing assistant to AI as a workflow engine, which is the whole premise of the [AI Agent Series](https://www.appliedcomms.ai/tag/ai-agents-series/) we have been publishing this spring. If looking carefully, you can also see a shift from my natural cynicism to becoming more of an advocate for AI's capabilities – while still wanting to draw attention to the challenges, shortcomings, and ethical issues.
So if the tools are this capable, why does so much AI in communications still feel stuck?
## **The bottleneck moved**
Here is the year's most important lesson, and it runs through almost every interview in the series. The limiting factor is no longer the model (at least when it comes to [those at the frontier](https://www.nvidia.com/en-us/glossary/frontier-models/)). It is the conditions around it: the state of an organisation's knowledge, its honesty, and its willingness to change how it works.
[Neal Mann, the NOAN founder](https://www.appliedcomms.ai/noan-neal-mann-interview/), put the knowledge problem most bluntly. "You cannot put AI on top of a mess of business knowledge and expect accurate results," he told us. In his analysis, hallucinations are not a bug to be patched but a structural feature of any system asked to reason over contradictory inputs: 25 versions of the brand positioning, pricing that disagrees with itself, strategy buried in a deck nobody can find. The model is fine. The inputs are a mess, and that is an organisational problem wearing a technical disguise.
[Ben Verinder, eight years into researching AI in PR](https://www.appliedcomms.ai/ben-verinder-ai-pr/), found a second wall, and it is about trust. His data suggests around half of in-house teams are not asking their agencies anything about AI use at all, and agencies report the mirror image. He points to research showing a measurable trust penalty when stakeholders discover AI has been used in ways they would not have sanctioned. Disclosure on its own does not fix that; the use has to be socialised. His sharpest line is the one comms leaders most need to hear: "This is a change programme, not an IT programme."
Then there is the failure mode with nothing to do with capability at all. Almost every team that has tried AI has a pilot that worked: an agent that drafts the newsletter, a workflow that turns a report into a week of content in an afternoon. Six months later the pilot is still a pilot, or it has quietly stopped, because the one capable person who built it got busy or moved on. The tools were never the hard part. Building an organisation that can hold the new way of working after the novelty fades is the hard part, and it is the argument the AI Agent Series builds toward in its [closing phase](https://www.appliedcomms.ai/the-pilot-trap-comms-with-ai-in-the-transform-phase/).
**Put those three together and you have a real state of the nation. The demo is solved. The deployment is not. The gap between the two is where almost all the difficulty now lives, and it is made of trust, knowledge and organisational design, not tokens.**
## **What this means if you lead a comms team**
The temptation, reading the capability story, is to accelerate: do the same things, faster. The year's evidence says that is the trap. The teams getting value are the ones treating this as a redesign of how the work is done, not a speed upgrade.
Four things follow, and none of them are about choosing a tool.
1. Start from problems, not products. The teams Ben sees getting it right do not ask "what can we use AI for?" They ask "what problems do we have that AI could help with?"
2. Fix the knowledge layer before you scale AI on top of it, or you simply automate the mess faster.
3. Name ownership and a shared standard for what good AI-assisted work looks like, so a useful pilot becomes a capability rather than a story the team tells about the time it tried AI.
4. And invest in judgement over prompt-craft: the scarce skill is knowing when to trust an output and when to distrust it, which is the briefing discipline most communicators already have.
This is also, candidly (and with admitted self-interest), where the two halves of our own ecosystem earn their keep. The [Comms With AI templates](https://www.commswith.ai/) exist to make the repeatable parts repeatable. The harder, organisation-specific work, the readiness, the redesign, the change leadership, is what my digital comms consultancy [Faur](https://faur.site/) does hands-on. Neither lets you buy your way past the deployment gap. They are ways of doing the work the gap demands.
## **The year in one place**
If you have joined recently, much of what is above was published before you arrived. So here is the back catalogue in one place, the evidence this scorecard is built on, grouped by the kind of read each one is. It starts where the publication did, with [why Applied exists and what to expect](https://www.appliedcomms.ai/why-applied-comms-ai-and-what-to-expect-2/).
**The AI Agent Series.** The six-phase operating model for comms with AI ([series tag](https://www.appliedcomms.ai/tag/ai-agents-series/)):
- [The framework](https://www.appliedcomms.ai/ai-agent-driven-communications-practical-framework/)
- [Strategise: the work before the work](https://www.appliedcomms.ai/ai-agent-series-strategise/)
- [Create: the volume problem](https://www.appliedcomms.ai/comms-with-ai-in-the-create-phase/)
- [Govern: the cost of a miss](https://www.appliedcomms.ai/ai-agent-series-govern/)
- [Monitor: the lag problem](https://www.appliedcomms.ai/the-lag-problem-comms-with-ai-in-the-monitor-phase/)
- [Transform: the pilot trap](https://www.appliedcomms.ai/the-pilot-trap-comms-with-ai-in-the-transform-phase/)
**Interviews.** Practitioners on what AI is really doing to the work (many thanks again to all those who I spoke to!):
- [Barney Evison of Flipside, from Watson to Lovable](https://www.appliedcomms.ai/interview-flipside-barney-evison/)
- [Joyce Higgins of FleishmanHillard, the AI architect](https://www.appliedcomms.ai/interview-fleishmanhillard-joyce-higgins/)
- [Neal Mann of NOAN, the living fact layer and the death of the PDF](https://www.appliedcomms.ai/noan-neal-mann-interview/)
- [Rik Turner, the SEO-PR strategist who saw AI search coming](https://www.appliedcomms.ai/rik-turner-pr-for-ai-interview/)
- [Ben Verinder, the honesty gap](https://www.appliedcomms.ai/ben-verinder-ai-pr/)
**Builds and experiments.** Things I made, including what broke:
- [How we build: app development ground rules](https://www.appliedcomms.ai/app-building-ground-rules/)
- [A LinkedIn content creator, built with Claude](https://www.appliedcomms.ai/claude-artifact-app-builder-linkedin-content-creator/)
- [The perfect AI prompt for media pitch subject lines](https://www.appliedcomms.ai/ai-prompt-media-pitch-subject-line/)
- [A 'will my boss hate this?' detector, built with Lovable](https://www.appliedcomms.ai/ai-stakeholder-comms-review-tool/)
- [A voice-enabled AI strategy architect](https://www.appliedcomms.ai/voice-enabled-ai-strategy-architect-campaign-dashboard-google-gemini/)
- [Claude Cowork, from content creator to system builder](https://www.appliedcomms.ai/claude-cowork-review/)
- [Building a live AI webinar with the workflow I was teaching](https://www.appliedcomms.ai/ai-webinar-workflow/)
- [An ICP agent to build an AI target reader](https://www.appliedcomms.ai/icp-agent-ai-target-reader/)
- [Building Comms With AI with Claude Code](https://www.appliedcomms.ai/comms-with-ai-claude-code-build/)
**Deep dives and trends.** The thinking underneath the builds:
- [Prompt engineering: our production standards](https://www.appliedcomms.ai/prompt-engineering-production-standards/)
- [How project folders supercharged my AI comms workflow](https://www.appliedcomms.ai/project-folders-ai-comms-workflow/)
- [AI in crisis comms: help or hindrance?](https://www.appliedcomms.ai/crisis-comms-ai/)
- [2026 communications trends](https://www.appliedcomms.ai/2026-communications-trends/)
**Tool reviews.** Tested properly, not just tried:
- [Our tool-testing methodology](https://www.appliedcomms.ai/tool-testing-review-methodology/)
- [Testing ChatGPT's Deep Research](https://www.appliedcomms.ai/deep-research-comms-test/)
- [Grammarly: your everywhere comms subeditor](https://www.appliedcomms.ai/grammarly-review/)
- [A complete pitch suite in under an hour, with Claude Skills and Projects](https://www.appliedcomms.ai/claude-skills-projects-business-pitch/)
## **Year two**
Earlier this month, at the inaugural [AI Comms Awards](https://www.commswith.ai/updates/ai-comms-awards-2026-result/) run by Communicate Magazine, [Comms With AI](https://www.commswith.ai/), the free resource that grew out of this publication, won Gold for Best Innovation in AI Tools for Communications, and I was named AI Communications Leader of the Year. It was a hell of a way to bring in year two. So what now, you may ask?
Well, a note of caution to close, borrowed from Ben: anyone forecasting AI's path with certainty is "either overconfident or telling fibs, or both." So no predictions here.
What is safe to say is that the frontier for exploration has moved markedly. For the year ahead, the experiments worth running, and documenting, are less about whether the next model can do the task, and more about whether real teams, with real politics and real legacy processes, can put it to work without losing the judgement that made them good in the first place.
That was the promise a year ago: experiment, fail, learn, share. The experiments have only got more interesting. See you for year two.
## **Coming up**
If those year-two questions are the right ones, two things over the next fortnight put them to work, and you are welcome at both – in fact, I firmly encourage you to take a look and sign up/share with others:
- **This Wednesday, 1 July**, the next live Applied / Comms With AI Leader Interview: [Elif Güvençer on her Two Clocks framework](https://www.eventbrite.co.uk/e/the-two-clocks-elif-guvencer-on-repositioning-comms-for-the-ai-era-tickets-1991391312021), a conversation about repositioning comms for the AI era. It lands squarely on the question this piece keeps circling: what the function is actually for once the tools can do the doing. Free, online, and recorded if you cannot make it live.
- **The week after, on 7 and 8 July**, I am running two hands-on AI training sessions for comms teams with [Big Fish Training](https://www.bigfishtraining.com/articles/new-ai-training-courses-for-pr-professionals.cfm): one for account executives, one for account managers and directors. Not talks about AI, but the workflow running live, the kind of in-the-room work the templates were never meant to replace.
Both are the human layer this whole piece argues the tools cannot close on their own. Come and join in, and I guarantee you'll have several takeaways to put in practise.
---
*Applied / Comms With AI documents what actually works in AI for communicators, and what does not. Explore the free resource at* [*Comms With AI*](https://www.commswith.ai/)*, and subscribe at* [*appliedcomms.ai*](https://www.appliedcomms.ai/)*.*
### Comms With AI: How I Built an Award-Winning AI Communications Tool in 3 Months (and What I Got Wrong)
URL: https://www.appliedcomms.ai/comms-with-ai-how-i-built-an-award-winning-ai-communications-tool-in-3-months-and-what-i-got-wrong/
Last updated: 2026-06-22T07:44:42.000Z
*Three months ago I documented building* [*Comms With AI*](https://www.commswith.ai/) *with Claude Code. On 18 June it won Gold at the inaugural AI Comms Awards, and I was named AI Communications Leader of the Year. In between, and more quietly, it rebuilt itself from a template library into something closer to a product. Here is why I built it, what the rebuild taught me, and the calls I would make differently.*
In March (which now feels several lifetimes ago), [I wrote up how Comms With AI got built](https://www.appliedcomms.ai/comms-with-ai-claude-code-build/): roughly ten hours of Claude Code to a functional site, then two months of structured iteration to make it worth using. That piece ended on an open question. If the tools have made building this easy, what is the actual moat? My answer then was the professional judgement embedded in the content, and I said I would come back to it once the site had been in real use.
Fast forward to Thursday evening, and at the inaugural [AI Comms Awards](https://www.commswith.ai/updates/ai-comms-awards-2026-result/) on 18 June – run by the excellent [Communicate Magazine](https://www.communicatemagazine.com/) – we won **Gold for Best Innovation in AI Tools for Communications**, while I was named **AI Communications Leader of the Year** for the work behind it.
The judges' citation described the platform as "simple, specific and scalable", and one judge called it "an innovative tool grounded in strong communications use cases and delivering real results". As you'd imagine, I was blown away by the results – and keen to make the most of this result to further what I've been working on and sharing with others.
However, before anything else, I want to be upfront about what that does and does not mean, because practical honest is at the core of the Applied house style. An award does not validate the content. Only sustained real-world use can do that. What it does tell me is that a resource launched in weeks, then sculpted over a few months, by someone who is not a developer, can now stand next to established industry work and not look out of place. Eighteen months ago that sentence would have been a stretch.
That is a more interesting story than any bloated self-congratulatory puff piece, and so here I'll aim to delve into the what, why, and how of this – as well as what lies ahead. (While admittedly still patting myself on the back a wee bit...)
## **What Comms With AI is – and why I thought this needed to exist**
Let's start with the why, since I skated over this in March, and it is the part that actually explains the rest.
[Applied / Comms With AI ](https://www.appliedcomms.ai/)began life last summer as a newsletter, written while I was working through Imperial College's [Professional Certificate in Machine Learning and Artificial Intelligence](https://certificates.emeritus.org/profile/michaelmaclennan614244/wallet?ref=appliedcomms.ai). The frustration that started it was simple. A bulk of what was espoused about AI for communicators was either breathless evangelism or vendor marketing dressed up as insight. What was largely missing was honest, practical guidance from people who actually understand the work.
CommsWith.AI (as it was initially styled, thanks to a random domain purchase alongside the appliedcomms.ai get) grew straight out of that. The need I was building for was not "more AI content". It was a place where a comms professional could find something tested, specific to their job, and usable inside 5-10 minutes, without wading through hype to get to it.
Therefore, the first version of CWAI was deliberately concrete: 47 templates across seven workflow categories and six toolkits. The intent was utility. Take a prompt, get on with your day.
Comms With AI: the tool at launch
That intent was right, which is more than can be said about the organisation...
## **What I missed at the beginning**
The library now carries 62 templates, seven toolkits and a directory of 44 tool reviews, with new bundles such as the [Crisis Preparedness](https://www.commswith.ai/toolkits/crisis-preparedness/) and [Campaign Measurement](https://www.commswith.ai/toolkits/campaign-measurement-roi/) toolkits, and a [five-minute readiness diagnostic](https://www.commswith.ai/readiness/) that scores a comms function across six dimensions and points the visitor to a sensible starting point rather than a wall of templates. Crucially, we now have a functional beta of [Plan Comms With AI](https://www.commswith.ai/plan/), directly using AI to provide an additional layer of support.
Three months later, 5-10 hours per week of dedicated work. The rate of change surprised me, and I often balanced the exhiliration of this against the frustrations of still spending more hours than it felt worth struggling with setups in time-honoured tools such as Mailchimp and Canva (the more things change etc etc).
The resource count is not the headline change: it's that the seven flat categories are gone, replaced by [a five-phase operating system](https://www.commswith.ai/os/): Strategise, Create, Govern, Monitor, Transform.
If those five sound familiar, they should. They are the spine of the [AI Agent Series](https://www.appliedcomms.ai/tag/ai-agents-series/) I have been publishing this spring, named explicitly when the series [set out its framework in April](https://www.appliedcomms.ai/ai-agent-driven-communications-practical-framework/): the cycle that runs from the [work before the work](https://www.appliedcomms.ai/ai-agent-series-strategise/), through [creation](https://www.appliedcomms.ai/comms-with-ai-in-the-create-phase/) and [governance](https://www.appliedcomms.ai/ai-agent-series-govern/), into [monitoring](https://www.appliedcomms.ai/ai-agent-series-monitor/), and finally into [building the capability to do it better next time](https://www.appliedcomms.ai/ai-agent-series-transform/).
### Regrets, I've had a few...
One thing I would have done differently: the model was sitting in my own published thinking the whole time, and I still shipped seven sensible-looking categories instead. A flat taxonomy is fine at 47 items. At 62, with a clear path to many more, it fragmented: categories overlap\[ed, a visitor cannot tell where a new template liveed, and the thing stopped feeling like a system and started feeling like an unruly pile.
The phase model fixes that at the root. Every template has one obvious home, every toolkit maps to a stretch of the cycle, and the structure does work that no amount of individual template quality can do on its own. I watched the categories strain as the library grew, then spent real effort retrofitting an operating system onto content filed against the wrong logic. **The lesson is the one I would press hardest on anyone doing this: decide the fundamental structure before you scale, because retrofitting it later is the most avoidable work there is.**
## **From handy resource to award-winning product**
The other shift is harder to see in a screenshot. CWAI started as a handy resource, a place to grab a template and go. Three months on it behaves like a product. The diagnostic routes you. The operating system organises you. A new front door, [Plan Comms With AI](https://www.commswith.ai/plan/), now in open beta, turns a plain description of what you need to communicate into a brief, a route through the library, and a copy-ready prompt.
Around the free library there is now a set of paid options for the work the templates cannot do on your behalf: a [consultation booking option](https://www.commswith.ai/consult/), [training delivered alongside the wonderful Big Fish Training,](https://www.bigfishtraining.com/articles/book-now-ai-training-courses-for-pr-professionals-july-2026.cfm) and a [Deploy service](https://www.commswith.ai/deploy/) for governed AI rollout inside an organisation.
The publication arm has folded in alongside all of this. Applied is no longer a sister project running in parallel; it is the learning lab that feeds the platform, now including a live interview series. The first, [The Honesty Gap with Ben Verinder](https://www.appliedcomms.ai/ben-verinder-ai-pr/) on AI, PR and trust, set the format. The next, [Elif Güvençer on her Two Clocks framework](https://www.eventbrite.com/e/the-two-clocks-elif-guvencer-on-repositioning-comms-for-the-ai-era-tickets-1991391312021?aff=oddtdtcreator), runs on 1 July. And the first vertical, [Leader Comms With AI](https://www.commswith.ai/leader/), built for senior comms leaders rolling out AI across their teams, opens later this week with a free Leadership AI Governance Toolkit.
One principle set at launch held under all that pressure, and I have absolutely zero plans to ever change this. **The library is free and stays free. No gate, no account, no email wall on a single template.** The paid layers exist for the organisation-specific work, the readiness, the governance, the redesign, that a generic template was never going to solve. **Templates make the repeatable parts repeatable, and the hard, contextual work is done with people.**
It is the moat question from March, answered slightly differently. Defensibility was never going to be the templates, which anyone can now generate. It is the judgement about how they fit together, what good looks like, and what a given team should do next. The structure and the services are both expressions of that judgement, and I think that is what the award was actually recognising.
## **What I got wrong, and what I have not proven**
Four things:
- The structure call is the big one, and I have already made it, so I will keep it short: build the operating model before you scale, not after. What I would add now is why it matters beyond tidiness. comms people are not short of tools to read about. They are short of time. The reader I am building for is not someone with an afternoon to delve into every template; it is someone stretched thin, between deadlines, who needs to reach the one relevant thing quickly. Structure is what makes that possible. A coherent model is not housekeeping. It is the whole point.
- The second is scope. When building is this cheap, the temptation is to keep adding, and "more templates" feels like progress because it is measurable. It mostly is not. A few of the things I added early earned their place. A few were there because I could, and they diluted rather than strengthened. The discipline that matters now is subtraction – or at least maintaining a considered balance – and it is harder than building, because nothing prompts you to do it.
- The third I cannot resolve yet. I still do not have a confident read on how the resource performs in genuine, sustained use, as opposed to first visits and kind words. That was the open question in March, and the award does not close it. Usage data is starting to accumulate, and I will report it further when there is a scope of time which says something real rather than something flattering. The honest position three months in is that the build is proven and the value in the wild is not, quite, yet.
- Finally, something which deserves its own section...
## **The part that needs people**
There is one more lesson, and it came from outside the glare of the screen.
At the [AI for PR Conference](https://www.communicatemagazine.com/conference/ai-for-pr-conference-2026/), which sold out with almost 400 attendees and is already confirmed to return in 2027, the questions that stayed with me after were not about prompts or tools. They were about roles. Practitioners asking, in different ways, how their work with clients changes when the client can do more of it themselves. It was a consistently aired and fair worry: from my own experience, **I think that as in-house teams take on more of the doing, the agency relationship moves further towards advice and consulting. That is exactly where most of our experience actually lives, and it is increasingly what I am building the platform around.**
Training is part of the same realisation. I barely thought about it at the start, and it is now becoming central. Handing someone a good prompt is useful, but it does not replace a proper conversation, whether that is between a comms agency and its client, or me running a hands-on session, or one of the Leader webinars. Building those human layers back into what started as a pure resource bank has been the most rewarding part of the last three months, and the least expected.
## **What this means if you are building too**
The practical takeaway is not the toolchain, which will be different by the autumn. It is that as the tools get easier, the fundamentals get more decisive, not less. A coherent model of the work beats a fast accumulation of parts. Knowing what *not* to build beats knowing how to build it. And the questions worth asking before any of it are not "what can I make?" but "what should this be, who is it really for, and how should it be organised so it still makes sense at ten times the size?"
The two awards are a welcome marker, and I will not pretend otherwise – not least since they'll be appearing in the background of any Zoom call from today onwards... However, the rebuild around a stronger structure is for me the key lesson. It's what I primarily got wrong when first launching, and I am keen to see how things now perform with this in place. I will take both into the next period of CWAI and report back, as ever, on what works and what does not.
## **Where we go next**
The library stays free, and the next three months are about adding the human layers around it.
On 1 July, the interview series continues with [Elif Güvençer on her Two Clocks framework](https://www.eventbrite.com/e/the-two-clocks-elif-guvencer-on-repositioning-comms-for-the-ai-era-tickets-1991391312021?aff=oddtdtcreator), a conversation about repositioning comms for the AI era. It is free to join, and shall be a fascinating chat.
And on 7 and 8 July, I am running two hands-on AI training sessions for comms teams with [Big Fish Training](https://www.bigfishtraining.com/articles/book-now-ai-training-courses-for-pr-professionals-july-2026.cfm): one for account executives, one for account managers and directors. These are the proper conversations the templates were never meant to replace!
And last but certainly not least, this week the first vertical goes live: [Leader Comms With AI](https://www.commswith.ai/leader/), built for senior comms leaders rolling out AI across their teams. It opens with a free Leadership AI Governance Toolkit, the part most teams are missing.
If any of that is useful, come and join in. This is what [Comms With AI ](https://www.commswith.ai/)is all for.
---
*Applied / Comms With AI documents what actually works in AI implementation for communications professionals, including what does not. Explore the resource at* [*CommsWith.AI*](https://www.commswith.ai/)*, and subscribe at* [*appliedcomms.ai*](https://www.appliedcomms.ai/)*.*
### The Pilot Trap: Comms With AI in the Transform Phase
URL: https://www.appliedcomms.ai/the-pilot-trap-comms-with-ai-in-the-transform-phase/
Last updated: 2026-06-16T06:00:24.000Z
Almost every communications team that has tried AI has a pilot that worked.
Someone, usually one capable and curious person, built something. An agent that drafts the monthly newsletter. A workflow that turns a report into a week of social content in an afternoon. A monitoring brief that used to take half a day and now takes twenty minutes. It worked. People were impressed. There was a meeting where someone said, 'We should do more of this.
And then, often, nothing. Six months later, the pilot is still a pilot, or it has quietly stopped, because the person who built it got busy, or moved on, or simply could not carry it alone. The capability never spread. The impressive thing became a story that the team tells about that time they tried AI.
This is the pilot trap, and it is the most common way AI fails in communications. Not with a dramatic mistake. With a slow fade. **The tools were never the hard part. The hard part is building an organisation that can hold the new way of working after the novelty wears off**, and that is the Transform phase of the Comms With AI Operating System.
If you have followed the series this far, you have seen the four phases that produce and protect the work: [Strategise](https://www.appliedcomms.ai/ai-agent-series-strategise/), [Create](https://www.appliedcomms.ai/ai-agent-series-create/), [Govern](https://www.appliedcomms.ai/ai-agent-series-govern/), and [Monitor](https://www.appliedcomms.ai/ai-agent-series-monitor/). The [overview introduced all five](https://www.appliedcomms.ai/ai-agent-driven-communications-practical-framework/), and the [full series is here](https://www.appliedcomms.ai/tag/ai-agents-series/). This final article is about the phase that decides whether any of it lasts.
---
## **The Transformation Problem**
Here is the distinction the pilot trap turns on. A pilot that depends on one enthusiastic person is not a capability. It is a hobby that happens to be useful. The moment that person is unavailable, the capability is unavailable, and an organisation cannot plan around a hobby.
Transform is the work of converting hobbies into capabilities. And the reason it is hard, the reason teams skip it, is that it is not technical work. Deploying an agent is comparatively easy. Changing how an organisation operates is not, because it touches the things organisations find genuinely difficult: roles, skills, incentives, ownership, and culture.
Consider what actually has to change for AI-assisted communications to become normal rather than novel. Job descriptions have to acknowledge it, or it stays off-the-record. People need new skills, and not the ones usually advertised. Someone has to own the decision to adopt, maintain, and retire tools, or the organisation accumulates a drawer of half-used subscriptions. There has to be a shared standard for what good AI-assisted work looks like, or every person applies their own. And the culture has to make it safe to say a workflow is not working, or problems go quiet instead of getting fixed.
None of that is a tooling problem. All of it is a leadership and organisational problem. That is why Transform is the phase where the AI does the least, and the phase that most determines whether the other four were worth the effort.
---
## **What the Transform Phase Covers**
Transform is the capability layer of the Operating System. It holds the work of building an organisation that can sustain and improve agent-driven communications over time:
- [**AI readiness assessment**](https://www.commswith.ai/readiness/): an honest diagnostic of where the team stands across all five phases
- **Workflow redesign**: rebuilding processes around what agents can now do, rather than bolting agents onto old processes
- **Role evolution and team structure**: how communications jobs change, and what the team needs to look like
- **Training and upskilling**: building the specific judgement that AI-assisted work requires
- **Tool evaluation and selection**: choosing, maintaining, and retiring tools deliberately
- **Change management**: leading people through the shift, not just announcing it
- **Operating model design**: deciding where agents live, who maintains them, and how the whole system is governed
Unlike the other phases of the Operating System, the [Transform library on Comms With AI](https://www.commswith.ai/library/transform/) is deliberately small: four templates covering the structured, diagnostic end of the phase. The [AI Readiness Assessment](https://www.commswith.ai/library/transform/ai-readiness-assessment/), [Comms Workflow Audit](https://www.commswith.ai/library/transform/comms-workflow-audit/), [Capability Gap Analysis](https://www.commswith.ai/library/transform/capability-gap-analysis/) and [AI Tool Evaluation Framework](https://www.commswith.ai/library/transform/ai-tool-evaluation-framework/) will get you a long way into the analytical groundwork. The rest of the phase is deliberately not templated. A template structures a repeatable task, and change leadership is not a repeatable task. It is leadership, judgement, and change work, specific to each organisation, and it is served through training and hands-on consulting rather than through a worksheet.
---
## **The 20/80 Rule for Transform**
Across the Operating System, the working rule is roughly 30% AI, 70% human. Transform sits at the human end of the range, at **20% AI, 80% human**, the same balance as Strategise and the lowest agent share in the framework.
This ratio shifts across the five phases:
- **Strategise: 20/80** – research and synthesis AI, judgement human
- **Create: 35/65** – drafting AI, voice and nuance human
- **Govern: 25/75** – systematic checking AI, reputational judgement human
- **Monitor: 40/60** – pattern detection AI, interpretation human
- **Transform: 20/80** – readiness assessment AI, change leadership human
There is a neat symmetry in the framework here. The two phases where AI does the least, Strategise and Transform, are the two that bookend the cycle: the thinking before the work, and the organisational learning after it. The phases in the middle, where work is produced and processed, are where agents carry more. AI is strongest in the doing and weakest at the edges, where judgement and leadership live. That is not a limitation to apologise for. It is the shape of the tool, and designing around it honestly is the whole point of the Operating System.
The 20% an agent contributes to Transform is real and worth having. It can run a structured readiness assessment, benchmark a team against a maturity model, map current workflows, model redesigned ones, and compare tools against defined criteria. That is genuine analytical help. But it cannot do the 80%, because the 80% is convincing people, navigating politics, building trust, holding a standard, and leading through the discomfort of changing how work is done. No agent leads change. People do, or it does not happen.
---
## **Where Agents Change Transformation Work**
Four applications are worth examining, though the honest framing for this phase is different from the others. In Strategise, Create, Govern, and Monitor, the agent does a meaningful share of the work. In Transform, the agent informs the work, and a human does almost all of it.
### **1\. AI readiness assessment**
The most useful starting application. An [agent-supported diagnostic ](https://www.commswith.ai/readiness/)works systematically through the five phases of the Operating System and builds an honest picture of where a team stands: where it is genuinely capable, where it is improvising, where it has tools without process or process without tools, and where the gaps are most expensive.
The value is in the structure and the honesty. Teams assessing their own AI maturity tend to anchor on their best example, the one impressive pilot, and assume it represents the whole. A structured assessment pushes past the showcase and looks at the routine. The output is not a verdict. It is a map of where to spend effort first, which is the most useful thing to have at the start of any transformation.
This site's own origin is a relevant example of readiness in action. The [build of the Comms With AI resource using Claude Code](https://www.appliedcomms.ai/comms-with-ai-claude-code-build/) was, in part, a practical readiness exercise: finding out by doing what was actually possible, where the limits were, and what skills the work demanded.
### **2\. Workflow redesign**
The instinct when adopting agents is to bolt them onto the existing process. The team keeps its current workflow and inserts an agent at one step. This produces a modest gain and misses most of the value.
Real workflow redesign asks a harder question: if we were designing this process now, knowing what agents can do, what would it look like? An agent is useful for mapping the current workflow in detail and modelling redesigned versions, but the redesign decisions, what to keep, what to cut, what to resequence, are human and often political, because workflows encode who does what and changing them changes that.
### **3\. Tool evaluation and selection**
Communications teams accumulate tools the way garages accumulate tins of paint. A structured evaluation, comparing tools against the team's actual needs rather than their marketing, is work an agent can scaffold well.
Applied / Comms With AI has effectively been running this discipline in public since it started, through its [tool review methodology](https://www.appliedcomms.ai/tool-testing-review-methodology/) and the reviews built on it, and the [AI Tool Evaluation Framework template](https://www.commswith.ai/library/transform/ai-tool-evaluation-framework/) turns that method into something your team can run itself. The lesson from that work is consistent. The right tool is the one that fits the team's real workflow and skill level, not the one with the most features or the loudest launch. Transform is where a team decides to choose tools that way on purpose, and decides who owns the choice.
### **4\. Skills mapping and role evolution**
This is the application that matters most, because it is about people, and people are what make a capability last.
An agent can help map a team's current skills against the skills agent-assisted communications requires, and the gap is usually not where people expect. The scarce skill is not prompt-writing. It is judgement: knowing when to trust an agent's output and when to distrust it, when a draft is genuinely good and when it is merely fluent, when to push back. That judgement is built through guided practice, not a training course, and designing that practice is human work.
Underneath the skills question is a deeper one about what the job becomes. This site has [written before about the shift from content creator to system builder](https://www.appliedcomms.ai/claude-cowork-review/): the communications professional moving from producing every piece by hand toward designing, briefing, and supervising the systems that produce them. That is the central role change of the Transform phase. It is not a smaller job. The craft simply relocates, from the output to the system that generates the output, and from doing the work to being accountable for the work an agent did. A team that does not name and support that shift will find its best people either quietly resisting it or quietly leaving.
---
## **A Test Your Organisation Should Pass**
Run your organisation against the following five questions. They distinguish a real AI capability from a collection of clever pilots waiting to fade.
**1\. If your most AI-fluent person left tomorrow, would the capability survive?**
If the answer is no, you have a person, not a capability. That is the single most important question in the phase, and for most teams the answer is uncomfortable.
**2\. Is AI use written into how you describe roles, or is it off-the-record?**
If agent-assisted work is something people do but no job description mentions, it is unsupported by definition. What is not named is not resourced, not developed, and not defended.
**3\. Does your team share a standard for what good AI-assisted work looks like?**
Without a shared standard, every person applies their own, quality varies invisibly, and the Govern phase has nothing consistent to govern against. The standard does not need to be elaborate. It needs to exist and be agreed.
**4\. Who owns the decision to adopt, change, or retire a tool?**
If the answer is "nobody, really", the team will drift into tool sprawl, paying for things it half-uses and never deciding to stop. Ownership of the toolset is a named job or it is a mess.
**5\. Is there a real route for the team to say a workflow is not working, and be heard?**
If raising a problem feels like criticising the AI strategy, problems will go quiet rather than get fixed. A capability that cannot hear bad news cannot improve, and a capability that cannot improve will not last.
If your organisation passes all five, you have built something durable. If it fails any of them, that is the Transform work still to do, and it is the work that protects everything the other four phases produced.
---
## **Where To Start**
If Transform is the phase your team has skipped, which is the single most common gap in the whole Operating System, because it is the least visible and the least exciting, three starting points will move you furthest.
**Start with a readiness assessment.** Before redesigning anything, get a clear, structured, unflattering picture of where the team genuinely stands across all five phases. The [AI Readiness Assessment template](https://www.commswith.ai/library/transform/ai-readiness-assessment/) gives you the structure to do it in a working session rather than a quarter – or you can directly take [our free online assessment tool](https://www.commswith.ai/readiness/). Effort spent in the wrong phase is effort wasted, and most teams do not actually know where their gaps are until they look properly.
**Make one role's AI responsibilities explicit.** Pick a role and write agent supervision, workflow ownership, or quality-standard accountability into it formally. One properly named responsibility does more for durability than five informal enthusiasms, because it converts a hobby into a job.
**Build a shared quality standard before you scale.** Agree, as a team, what good AI-assisted work looks like, and write it down. Scaling without a shared standard does not multiply quality. It multiplies inconsistency, and it leaves Govern with nothing stable to check against.
The [Transform templates](https://www.commswith.ai/library/transform/) cover the diagnostic groundwork: readiness, workflows, capability gaps, tool evaluation. The change work they point to cannot be templated, because it is not a repeatable task. It is leadership work, specific to each organisation, and it is the work that decides whether an AI initiative becomes part of how the team operates or a story about something it once tried.
---
## **Closing the Loop: The Series in One Idea**
Six articles, five phases, one argument.
The argument was never that AI writes communications. It is that AI is a workflow engine, and that communications work, being high-volume, structured, repetitive, and high-stakes, is unusually well-suited to being run as a set of designed workflows rather than a stream of improvisation.
The Comms With AI Operating System is the map of those workflows: [Strategise](https://www.appliedcomms.ai/ai-agent-series-strategise/) to think clearly before the work, [Create](https://www.appliedcomms.ai/ai-agent-series-create/) to produce at the volume modern communications demands, [Govern](https://www.appliedcomms.ai/ai-agent-series-govern/) to protect the organisation from what it publishes, [Monitor](https://www.appliedcomms.ai/ai-agent-series-monitor/) to read the environment and shorten the lag, and Transform to build the organisation that can hold all of it.
The phases form a cycle, not a checklist, and Transform is where the cycle closes. The organisational learning of Transform feeds straight back into Strategise. A team that has been through the full loop once re-enters it sharper: clearer about where agents help, more honest about where they do not, and better at the judgement that the whole system still rests on. The AI/human ratio across the phases, from 20/80 in the thinking work to 40/60 in the monitoring, was the recurring reminder of that point. The human share never disappears. It concentrates on the decisions where experience is load-bearing, and it stays the majority of the work everywhere.
If there is one honest line to end the series on, it is this. The teams that win with AI in communications are not the ones with the most tools or the fastest output. They are the ones that treated this as a redesign of how the work is done, did the unglamorous Transform work of making it stick, and kept human judgement firmly in charge of the things that matter. That has been the whole series, and it remains the whole point.
---
## **Join the Discussion**
Two live sessions this summer pick up exactly where this article leaves off.
On **Wednesday 1 July** I am hosting the next live Applied / Comms With AI Leader Interview, [The Two Clocks: Elif Güvençer on repositioning comms for the AI era](https://www.eventbrite.co.uk/e/the-two-clocks-elif-guvencer-on-repositioning-comms-for-the-ai-era-tickets-1991391312021). Elif's framework names the tension at the heart of the Transform phase: one clock is the daily work, made faster; the other is the slower, structural job of redefining what the comms function is for. It is free, online (12:00 BST) and recorded. [Register here](https://www.eventbrite.co.uk/e/the-two-clocks-elif-guvencer-on-repositioning-comms-for-the-ai-era-tickets-1991391312021); if you cannot make the slot, sign up anyway and we will send the recording and write-up.
And if training is the Transform work your team needs, in July I am running two live online AI courses with Big Fish Training, with individual places open for the first time: **AI for Account Executives** on Tuesday 7 July and **AI for Account Managers & Directors** on Wednesday 8 July, both 9.30am–1pm, £249 + VAT per place. Practical, hands-on, and built around real comms workflows. [Full details and booking are with Big Fish](https://www.bigfishtraining.com/articles/book-now-ai-training-courses-for-pr-professionals-july-2026.cfm).
---
## **What Comes Next**
This is the final article in the AI Agent series, but it is not the end of the work. Applied / Comms With AI continues to document AI-assisted communications in practice: the experiments, the tool reviews, the interviews with practitioners doing this for real, and the honest accounts of what works and what does not. If the series was useful, [the full archive is here](https://www.appliedcomms.ai/), and the [Operating System templates on Comms With AI](https://www.commswith.ai/) are free to use across all five phases.
If your team is working on Transform, or on any phase of the Operating System, and you want to compare approaches, I am always happy to hear from fellow practitioners.
## Sign up for Applied / Comms With AI
The practical guide for communications leaders navigating AI
Subscribe
Email sent! Check your inbox to complete your signup.
No spam. Unsubscribe anytime.
---
## **About**
[Applied / Comms With AI](https://appliedcomms.ai/) is the practical guide for communications leaders navigating AI, grounded in hands-on experimentation, workflow transformation and real-world implementation. Read the full [AI Agent series here](https://appliedcomms.ai/tag/ai-agents-series/). [Comms With AI](https://www.commswith.ai/) is the companion template and resource library for communications professionals using AI; it holds the Operating System and its templates across all five phases, with the deeper Transform change work served through training and consulting, because organisational change is not a repeatable task. Both sit alongside [Faur](https://faur.site/), a communications consultancy pioneering practical AI expertise for organisations ready to implement at scale. The Transform phase is where Faur does its deepest work: readiness assessments, workflow redesign, team training and the full Operating System diagnostic. If your organisation needs hands-on help making AI-assisted communications stick, get in touch at michael@faur.site or connect with me on [LinkedIn](https://www.linkedin.com/in/michaelmaclennan).
---
*This article concludes the* [*AI Agent series published on Applied / Comms With AI*](https://www.appliedcomms.ai/tag/ai-agents-series/)*. The series maps to the* [*Comms With AI Operating System*](https://www.commswith.ai/start-here/)*: Strategise, Create, Govern, Monitor, Transform.*
### The Lag Problem: Comms With AI in the Monitor Phase
URL: https://www.appliedcomms.ai/the-lag-problem-comms-with-ai-in-the-monitor-phase/
Last updated: 2026-06-10T06:00:50.000Z
Picture your typical Monday coverage report.
It arrives at the start of the week and summarises what was said about the organisation over the previous seven days: the coverage, the mentions, a sentiment read, a few competitor notes. Someone spent real time assembling it. It is thorough, it is tidy, and it is a history lesson. Everything in it has already happened. The story that mattered most broke on Wednesday, and the window to shape it closed on Thursday, two working days before the report that mentioned it landed.
This is the lag problem: the gap between something happening and the communications team knowing enough to act. Most teams monitor in the rear-view mirror, learning accurately and late what the road behind them looked like. This of course can be and often is incredibly useful, but the Monitor phase of the [Comms With AI Operating System](https://appliedcomms.ai/ai-agent-driven-communications-practical-framework/) asks a different question: how small can that gap be?
> The real prize of agent-assisted monitoring is not a cheaper coverage report. It is a shorter distance between signal and response.
---
New to the series? The [overview introduces the Operating System and its five phases](https://appliedcomms.ai/ai-agent-driven-communications-practical-framework/), and the earlier articles cover [Strategise](https://appliedcomms.ai/ai-agent-series-strategise/), [Create](https://appliedcomms.ai/ai-agent-series-create/) and [Govern](https://appliedcomms.ai/ai-agent-series-govern/). You can also [read the full series here](https://appliedcomms.ai/tag/ai-agents-series/). This fifth phase reads the environment and feeds everything it learns back into the first.
---
# The Lag Problem
Traditional monitoring can fail comms teams in three connected ways, and agents address each one differently:
- The first is **time**. The insight arrives after the moment to use it. A weekly report is a structural decision to be several days behind the conversation. For a settled organisation in a quiet sector, that is survivable. For anyone exposed to a fast news cycle, it means the team routinely learns about what matters too late to respond well.
- The second is **volume without signal**. Most teams are not short of monitoring data; they have alerts, listening tools and dashboards producing a constant stream of it. What they lack is the human time to turn that stream into the few things that actually matter. The data scales effortlessly. The interpretation does not. So teams either drown, scanning everything and absorbing nothing, or quietly ignore most of the feed and hope the important item was not in the part they skipped.
- The third is **the dead end**. Even timely, sharp insight often goes nowhere. It lands in a report, the report is read or it is not, and the loop closes there. What the team learned rarely reaches the people doing the strategic work, where it could change a decision.
**Most monitoring failures are not data failures; they are time, signal and feedback failures.** Agents change the economics of all three, which is what makes Monitor the phase where AI earns its highest share of the work.
---
# What the Monitor Phase Covers
Monitor is the intelligence layer of [the CWAI Operating System](https://www.commswith.ai/os/). It holds everything involved in reading the external and internal environment and turning it into something the organisation can act on:
- **Media and social monitoring**: tracking coverage, mentions and conversation across owned, earned and social channels
- **Issue tracking and escalation**: spotting emerging issues, judging trajectory, deciding what needs attention
- **Sentiment and narrative analysis**: not just whether the tone is positive or negative, but which narratives are forming
- **Campaign performance measurement**: tracking whether communications work is doing what it was designed to do
- **Competitive and stakeholder intelligence**: watching how others in the field are positioning and moving
- **Reporting and briefing**: turning all of the above into something a busy leader can absorb in two minutes
The phase before this one, Govern, protects the organisation from what it publishes. Monitor protects it from what it has not noticed.
---
# The 40/60 Rule for Monitor
Across the Operating System, the working rule is roughly 30% AI, 70% human. Monitor is the one phase where AI does more than the average, at **40% AI, 60% human**, the highest AI share in the framework.
This ratio shifts across the five phases:
- **Strategise: 20/80** – research and synthesis AI, judgement human
- **Create: 35/65** – drafting AI, voice and nuance human
- **Govern: 25/75** – systematic checking AI, reputational judgement human
- **Monitor: 40/60** – pattern detection AI, interpretation human
- **Transform: 20/80** – readiness assessment AI, change leadership human
Monitor tilts furthest toward AI because the core task suits agents almost perfectly. It is high-volume, continuous, pattern-detection work that rewards tirelessness and consistency. No communications professional can watch every channel without fatigue. An agent can.
But 40/60 still leaves the majority of the work with people. **An agent can tell you that mentions are up, that a phrase is recurring, that sentiment has shifted. It cannot reliably tell you whether that matters.** Unless the system has been set up to weight sources properly, it may not know that a recurring phrase came from one influential account rather than two hundred irrelevant ones, or that a sentiment dip is the predictable noise around a product update rather than a genuine problem. And it cannot make the escalation call, because escalation is a judgement about consequences, and consequences depend on context the agent does not have.
So the division in Monitor is unusually clean. Agents do the watching. Humans do the meaning.
## Sign up for Applied / Comms With AI
The practical guide for communications leaders navigating AI
Subscribe
Email sent! Check your inbox to complete your signup.
No spam. Unsubscribe anytime.
---
# Where Agents Change Monitoring Work
Four applications are where the change is most tangible:
## 1\. Always-on detection and anomaly alerting
This attacks the lag problem directly. Instead of a person assembling a report on a schedule, an agent watches continuously and stays quiet until something is worth saying. It learns the normal pattern, the usual volume of mentions, the usual mix of channels, the usual sentiment range, then flags departures from it: a sudden rise in mentions, a sentiment movement outside the normal band, a new account entering the conversation with reach, a phrase appearing repeatedly that was not there yesterday.
The shift is from a report to an alert. A report tells you what happened on a timetable; an alert tells you something is happening now, while you can still do something about it. For any organisation exposed to fast-moving risk, this is the single most valuable thing the phase offers.
Two potential hitches here. The first is calibration: an agent set too sensitive becomes a smoke alarm that goes off when you make toast, and a team that has learned to ignore its alerts is worse off than one with no alerts at all. Finding that balance is a matter of judgement, relying on your expertise. The second is source quality. The first monitoring design question is not which model to use; it is which sources matter enough to watch. An agent reading the wrong feeds will confidently miss the right story – much as we ourselves often do nowadays in a media ecosystem drowning in poorly generated slop.
## 2\. Sentiment and narrative analysis
Sentiment scoring, on its own, is a crude instrument – surely it's not just me continuously frustrated by it? "62% positive" feels informative and rarely is; it collapses a complicated conversation into a single dial and tells you almost nothing about why.
Narrative analysis is the more useful application. Rather than scoring tone, an agent reads across a body of coverage and identifies the narratives forming inside it. Not "sentiment is down" but "a specific framing of this issue is gaining traction, it started in one place, and it is being picked up in these terms." That tells you what story is being told about you, who is telling it and whether it is spreading, the thing a communications leader actually needs in order to decide whether and how to respond. It connects directly to the shift toward [AI-mediated search and visibility that Rik Turner described on this site](https://appliedcomms.ai/rik-turner-pr-for-ai-interview/). What stays human is the judgement about which narrative is a genuine threat and which is background noise that will fade on its own.
## 3\. Briefing automation
This gives a team the most time back. The work of turning raw monitoring data into a readable brief is real, skilled and almost entirely automatable. An agent can take the feeds and produce a short daily or weekly brief in a consistent structure: what changed since the last brief, what it appears to mean, and what, if anything, needs a decision. A human reviews it, corrects the interpretation, adds the context the agent could not have, and sends it. The brief that used to take half a day takes twenty minutes to check.
This is territory the site has tested before: the earlier review of [ChatGPT's deep research capability](https://appliedcomms.ai/deep-research-comms-test/) was, in effect, an examination of how well an agent does this kind of synthesis-and-summary work, and where it still needs a human hand. The [Weekly Monitoring Brief](https://www.commswith.ai/library/monitoring/weekly-monitoring-brief/) and [Monthly Stakeholder Update](https://www.commswith.ai/library/monitoring/monthly-stakeholder-update/) templates on CWAI give that output a structure the team can rely on, so the brief reads the same way every time.
## 4\. Predictive and early-warning intelligence
This is the application with the most promise and the most need for honesty.
The genuine version is real and useful. By tracking the early shape of issues, an agent can flag that something is on a trajectory that has, in the past, led to escalation: a small concern raised in more places, by more credible voices, in increasingly specific terms. That pattern often does precede a bigger story, and surfacing it early buys the team the rarest thing in communications, which is time. A signal is only useful, though, if the team knows what level of response it triggers, which is why early warning has to be paired with clear escalation thresholds, the job the [Issue Log Tracker](https://www.commswith.ai/library/monitoring/issue-log-tracker/) is built to do.
The overstated version is everywhere in vendor marketing and should be treated with suspicion. An agent does not predict the future; it extrapolates from patterns in the past. It is useful for the issue that follows a familiar curve and structurally blind to the genuinely novel event: the thing with no precedent in the data, which is very often [the one that does the most reputational damage](https://appliedcomms.ai/crisis-comms-ai/). "Predictive" is a fair description of trend extrapolation and a false one for foresight. Use the capability, value it, and never let it persuade the team that the unprecedented has been ruled out.
## What it looks like in practice
A charity launches a policy campaign on Monday. By Tuesday afternoon, an agent flags that a hostile framing is being repeated across three local outlets and one sector influencer, ahead of any coverage the team had clocked. They get the alert, update the FAQ, re-brief the spokesperson and adjust the paid social copy before the framing hardens into the dominant story. Monitor detected it, Govern checked the response, Create produced it, and the next Strategise cycle inherits a sharper read of how that issue travels. That is the loop working as designed.
---
# A Test Your Monitoring Should Pass
Run your current monitoring against these five questions. They separate monitoring that informs decisions from monitoring that documents the past.
1. **If something broke about your organisation right now, how long before you would know?** Be honest, and count in hours. If the answer is "the next report", you do not have monitoring. You have a record.
2. **Does your monitoring produce decisions, or only reports?** Look at your last month of output and find the decisions it changed. If you cannot point to any, it is running as an archive, not an intelligence function.
3. **Can you tell signal from noise, or are you tracking everything equally?** Monitoring everything is the same as monitoring nothing, because the one item that matters is buried in the thousand that do not. A working setup is mostly a definition of what counts.
4. **Does what you learn in Monitor actually reach the people doing Strategise?** If insight stops at a report and never reaches the people making strategic decisions, the loop is broken. The feedback path has to be deliberate; it does not happen on its own.
5. **What did your monitoring miss last quarter, and why?** Every setup misses things. A team that can answer this is tuning a system. A team that cannot is trusting one it has never tested.
Pass all five and it is genuinely an intelligence function. Fail any and the gap is in the Monitor phase, almost certainly costing you response time you cannot see.
---
# Where To Start
If Monitor is the phase your team has under-resourced, which is common, because it competes with the more visible work of producing things, start with three moves.
**First, replace one manual report with an agent-drafted, human-reviewed brief.** Choose the report that takes the most time and changes the fewest decisions. It is the fastest visible win in the phase and frees real capacity immediately.
**Second, set up anomaly alerting on your organisation's name, senior leaders, core products and live issues.** Keep the first version deliberately narrow, and expect to spend a few weeks tuning the sensitivity. A noisy system will not survive contact with a busy team. That tuning is the work, and it is worth it.
**Third, define the feedback route into Strategise.** Decide, explicitly, who sees monitoring insight, how often, and what kind of finding is allowed to change the plan. The best monitoring output is not the report itself. It is the change it causes in the next campaign brief, message house, stakeholder map or leadership decision. This is the step teams skip, and skipping it is what turns monitoring into an archive.
The templates in the Monitor phase are built around this logic: continuous agent detection feeding human interpretation, with the output structured so it informs decisions rather than filling folders. Start with the [Monitor templates on Comms With AI](https://www.commswith.ai/library/monitoring/): the [Weekly Monitoring Brief](https://www.commswith.ai/library/monitoring/weekly-monitoring-brief/), [Issue Log Tracker](https://www.commswith.ai/library/monitoring/issue-log-tracker/) and [Simple Comms Dashboard](https://www.commswith.ai/library/monitoring/simple-comms-dashboard/). (BTW If you need to design a monitoring system that feeds strategy rather than fills folders, [Faur](https://faur.site/) can help build the workflow, escalation logic and briefing structure around it. [Do get in touch](https://faur.site/contact)!)
---
# Join the Discussion
On **1 July** I am hosting the next live Applied / Comms With AI Leader Interview, [The Two Clocks: Elif Güvençer on repositioning comms for the AI era](https://www.eventbrite.co.uk/e/the-two-clocks-elif-guvencer-on-repositioning-comms-for-the-ai-era-tickets-1991391312021), a conversation about the structural choice facing every communications function as AI reshapes what the work is for, not just how fast it gets done. It is free, online and recorded. [Register here](https://www.eventbrite.co.uk/e/the-two-clocks-elif-guvencer-on-repositioning-comms-for-the-ai-era-tickets-1991391312021); if you cannot make the slot, sign up anyway and we will send the recording and write-up.
---
# What Comes Next
Article 6 is the final piece in the series, and it covers the **Transform** phase: building the communications organisation that can sustain all of this. The AI contribution drops back to 20/80, because Transform is not really about tools at all. It is about roles, skills, culture, and the difference between a workflow improvement that sticks and a pilot that quietly fades. It is also where the loop closes, because Transform feeds back into Strategise, and the cycle begins again.
If your team is working on Monitor-phase implementation and you want to compare approaches, I am always happy to hear from fellow practitioners at [michael@faur.site](mailto:michael@faur.site).
---
# About
[Applied / Comms With AI](https://appliedcomms.ai/) is the practical guide for communications leaders navigating AI, grounded in hands-on experimentation, workflow transformation and real-world implementation. Read the full [AI Agent series here](https://appliedcomms.ai/tag/ai-agents-series/). [Comms With AI](https://www.commswith.ai/) is the companion template and resource library for communications professionals using AI, and the [Monitor phase library](https://www.commswith.ai/library/monitoring/) covers media and social monitoring, issue tracking, sentiment and narrative analysis, briefing and stakeholder reporting. Both sit alongside [Faur](https://faur.site/), a communications consultancy pioneering practical AI expertise for organisations ready to implement at scale. If your team is working on the Monitor phase and needs bespoke support, whether monitoring design, early-warning systems or intelligence that feeds back into strategy, get in touch at [michael@faur.site](mailto:michael@faur.site) or connect with me on [LinkedIn](https://www.linkedin.com/in/michaelmaclennan).
---
*This article is part of the* [*AI Agent series published on Applied / Comms With AI*](https://appliedcomms.ai/tag/ai-agents-series/)*. The series maps to the* [*Comms With AI Operating System*](https://www.commswith.ai/start-here/)*: Strategise, Create, Govern, Monitor, Transform.*
### The Honesty Gap: Ben Verinder on AI, PR and Trust
URL: https://www.appliedcomms.ai/ben-verinder-ai-pr/
Last updated: 2026-06-02T04:00:23.000Z
*Ben Verinder has spent eight years researching what AI means for public relations, long enough to be wary of anyone who claims certainty about it. He is co-editor of* AI for Public Relations *, published this month by Kogan Page, an agency founder, a strategic communications adviser, and one of a small group of Founding Chartered PR practitioners globally. When Ben talks about AI, it is from evidence rather than enthusiasm. The argument he kept returning to in our conversation was not the policy gap the profession has learned to discuss, but a deeper one beneath it: an honesty gap, between teams and their agencies, and between communicators and the people they serve.*
---
This was the first live Applied Comms AI Leader Interview, and the format immediately proved its value: a serious conversation, sharp audience questions, and a guest willing to say plainly where the profession is still avoiding the hard parts. The full conversation [is available on YouTube](https://www.youtube.com/watch?v=dqzmW2Q%5F3EQ) and embedded below. What follows are the ideas that have stayed with me.
## Eight years in, and the generational divide is a myth
Ben's path into this was not post-ChatGPT opportunism. He was looking at machine learning in media and social monitoring platforms years before ChatGPT made AI a boardroom topic, working on a NATO project in 2015 and writing on AI and data ethics for the CIPR from 2017 onwards. Eight years in, the depth shows in the questions he chooses to take seriously.
I opened by asking Ben about UKRI's Press Officers' Day, where he had presented earlier that week, and which questions from the room he had not seen coming. He had fielded plenty of repeats, he said. The one that stuck came from a young attendee, who asked about generational differences in attitudes to AI.
The assumption in public relations, Ben explained, is that age produces considerable differences in how people feel about AI. It does not. He pointed to a survey from King's College London's new AI institute, covering 5,000 people, that found no meaningful age gap. The real differences sit elsewhere. There are some by gender, though more in adoption and confidence than in attitude: men tend to be overconfident about their training needs, women relatively underconfident. The sharper divide is by sector. Not-for-profit comms teams, Ben said, are less likely to be trained, less likely to have a team or organisational policy, more fearful of AI, and less confident it will improve their work.
He called that sector "a poster child for this inverse relationship between familiarity and fear". It is a useful reframing. Fear of AI is not mostly a matter of age or temperament. It tracks exposure. And closing that exposure gap, Ben argued, is squarely a job for communications.
## The gap everyone discusses, and the one beneath it
The policy gap is the part of this story the profession has learned to talk about. Ben's research points at something more uncomfortable underneath it.
His data suggests around half of in-house teams commissioning agencies are not asking those agencies anything about AI use at all. Agencies report the mirror image: roughly half say no client has ever asked. That figure has not shifted in two years. "There's a big gap in honesty and transparency around AI use," Ben said, "both from in-house teams and consultancies." It extends outward, too. Citing a 2025 Global Alliance study, he noted that comms teams are not being honest with their own stakeholders about AI use either.
Ben was careful not to overstate the profession's ethical maturity. The people most visible in AI discussions, he noted, are often those already connected to professional bodies, training and accreditation. That is not the whole market, and the picture in the wider, unregulated middle of the industry is likely worse than the one he sees from the inside.
There is a commercial reason this matters, and Ben thinks most teams have missed it. AI-generated content produced without sufficient human intervention attracts no copyright protection. So if an agency hands a client a body of largely machine-made work and assigns the copyright, that assignment, in Ben's words, "is not worth the paper it's written on", because there is no copyright to assign. Asking an agency how it uses AI is not box-ticking. It goes to the heart of the commercial agreement.
## The expectation has flipped
A year ago, the anxiety in comms was about being seen to use AI. Clients might think less of work they suspected was machine-made. That anxiety has not vanished, but Ben described a newer, sharper version of it. The default assumption has flipped: increasingly, stakeholders simply assume you are using AI.
That sounds like progress. Ben sees a trap in it. If everything you produce is assumed to be AI-generated, the human work loses its premium. His answer is to be specific in both directions. "One of the reasons you want to be really forthright about where you're using AI," he said, "is to demarcate where you're not."
Disclosure alone is not enough, though. Ben cited a University of Arizona study, built on 13 experiments and a substantial literature review, which found that using AI in ways stakeholders do not endorse carries a measurable trust penalty. Telling people you use AI does not protect you if they object to how. "We've got to socialise our AI use," he said. "It's not sufficient just to say, 'hey, we're doing this, guys.'" Getting that right is itself a public relations exercise, which, as Ben kept pointing out, makes it an opportunity for the people best equipped to do it.
## Two clocks, and the trouble with selling outputs
The commercial pressure on agencies came up repeatedly. Ben pointed to [Elif Güvençer's Two Clocks framework](https://www.linkedin.com/posts/elif-g%C3%BCven%C3%A7er-b3742349%5Fthe-clock-most-leaders-will-choose-and-why-activity-7461021783564128256-NNSH/), which captures the dual challenge well: an immediate clock for the daily work, and a structural clock for the harder work of redefining what the function is for and how it is measured.
Find The Two Clocks at reputationsignal.co
He summarised the logic with characteristic bluntness. If you sell outputs, you are in trouble. Press release distribution and output-based KPIs are exposed. If you sell outcomes, and genuine media relationships built on interpersonal trust with journalists, that is a different proposition: something AI can support but not replace. He would not, he said, sweeten that pill.
One story stuck with me. A freelancer was being hired by agencies to write tone-of-voice guidelines, which the agencies then used to train AI models, before telling her the AI could now do her job. The threat does not stop at the agency; it runs down the food chain. Ben helped her introduce a contract clause preventing her work being used to train AI. A small, practical win, and a reminder that some of the protection here is a drafting problem.
## A way in: the Bridges model
For teams that feel behind, Ben offered his BRIDGES model from the book (more about that[ in this other recent interview](https://www.vivapr.co.uk/the-truth-about-ai-with-ben-verinder/)). Its value is less in the acronym than in where it tells you to start. The teams getting this right, he said, are not asking "what can we use AI for?" They are asking "what problems or ambitions do we have that AI could help with?"
In practice, BRIDGES means briefing the right leaders first, starting from business problems rather than tools, running two or three bounded experiments, building prompt skill alongside them, agreeing basic team guidance, evaluating quickly, then supporting wider adoption.
The model briefs leadership and IT first, and Ben was firm that comms has to manage upwards here, because too many programmes are led by IT alone. "This is a change programme, not an IT programme." IT can govern the tools, he argued, but it cannot define the reputational judgement, the stakeholder acceptability or the tone of use. That is where communications has to lead.
He warned, too, against teams quietly self-funding licences or running shadow AI, (using AI in ways that are unsanctioned or tools that are prohibited) , which works only until it does not. From there: identify two or three tasks to experiment with, no more, because throwing AI at everything makes the impact impossible to measure. Develop prompt skills alongside. Put a basic team policy in place now rather than waiting for the perfect organisational one, which "is just right for about five seconds before the technology changes anyway".
Bad training creates its own drag. If a team's first AI session is generic, overhyped or irrelevant to their real work, Ben observed, they may be slower to come back when the next round has become genuinely useful. Evaluate honestly with a short huddle after a fortnight, then support the wider organisation.
The most memorable part was a story about shadow AI. When Ben trains a team, he runs an amnesty: get everyone in a room that trusts each other, and ask what they have actually been using AI for. "There hasn't been a session I've done in the last three years," he said, "where the head of comms hasn't gone, 'I did not know you were doing that.'" Sometimes it is faintly alarming. More often it is the opposite: a team realising it is further down the road than it thought.
He added a warning about champions. The person who self-selects as the AI lead may not be the one you would choose. He recalled a college chief executive whose trailblazers were the staff he would rather have kept in the classroom with students. It is an argument, Ben said, for strategic adoption over leaving it to whoever puts a hand up.
## What this means for comms leaders
For me, this was the clearest leadership lesson in the conversation. AI adoption does not begin with a tool rollout. It begins with finding out what people are already doing, deciding what is legitimate for the work and the stakeholders involved, and creating enough shared language that experimentation becomes visible rather than hidden. The risk is not only technical. It is reputational, commercial and cultural.
## The book that started on a dog walk
*AI for Public Relations* began, as Ben tells it, on a dog walk. He and co-editor Stephen Waddington were in various WhatsApp groups about AI, and worried that the tone of some was "a bit gung-ho". More seriously, they worried the profession lacked agency over how AI was reshaping it. "I don't want to belong to a club that's rubbish," Ben said. He wanted to gather as many informed voices as possible into one discussion. Waddington's answer was characteristically direct: write a book. If you build it, they will come.
They approached people they knew to be genuine experts and practitioners, among them Andrew Bruce Smith on technology, Richard Bagnall on measurement, Serena Mitchell on in-house adoption, Amy Mollett on AI policy, and Professor Anne Gregory and Dr Swati Virmani on AI's social ramifications. The book is deliberately practical, with key takeaways closing every chapter, and it stays away from forecasting the technology itself. The aim, Ben said, was something "timely, but not time-bound".
On what comes next, he was careful. Anyone forecasting AI's path with certainty, he said, is "either overconfident or telling fibs, or both". But he named trends he is confident about: pressure on early-career roles, changing team shapes, AI skills written into job descriptions, and a societal split in attitudes, with employers markedly more positive about AI's effect on jobs than the people they employ.
For comms specifically, he flagged AI-enabled deepfakes and disinformation as a core defensive skill the profession has to build. The point is not only that comms teams will use AI. They will operate in an AI-mediated information environment, where synthetic content, search summaries, deepfakes and stakeholder suspicion all affect trust. That is the strategic part of the conversation worth keeping in view, even when the immediate work is about tools and policies.
## One thing to do differently
I closed by asking what Ben would want a viewer to do differently after watching. His answer was unglamorous, and I think exactly right: learn to instruct an AI well. Even as agentic tools spread, understanding how to explain your workflow and your intent to a model is, in his words, "the singular most impactful thing you can do for yourself".
For comms teams, that means treating AI instruction as a briefing discipline: context, objective, audience, constraints, evidence, tone, review criteria. The skills are not foreign. They are the briefing craft most of us already practise, made explicit and made repeatable.
Ben practises what he preaches in how he stays current. Half a day a week, a Google Alert on artificial intelligence, every relevant story read and filed in a library tool called [Raindrop.io](http://raindrop.io). He deliberately does not use AI to summarise any of it. "Otherwise I don't learn anything myself." From one of the most rigorous voices in this field, that felt like just the right place to stop – leaving us with plenty to mull over and put into our own practice.
---
*AI for Public Relations: A How-To Guide for Implementation and Management*, co-edited by Ben Verinder and Stephen Waddington, is published by Kogan Page and[ available now](https://www.koganpage.com/marketing-communications/ai-for-public-relations-9781398625037). If you want the depth behind this conversation, start there. You can connect with Ben on[ LinkedIn](https://www.linkedin.com/in/ben-verinder/).
---
**Putting it into practice, live this Friday!** Ben's point about learning to instruct AI well is exactly what we will be doing in the open. On Friday 5 June, 13:00 to 14:00 BST, I am joining Emma Ewing of Big Fish Training for a free lunch-and-learn. We will take a vague, slightly chaotic comms brief and build it into a clear, structured campaign pack, live on screen, using a repeatable workflow in Claude (Projects and Skills). You can shape the session as it runs by suggesting audiences, objectives or curveballs of your own, with time for questions throughout. The focus is not on perfect outputs but on understanding where AI fits in real comms work, and where human judgement and quality control still matter. [Register here](https://us02web.zoom.us/webinar/register/WN%5FSnAfg2o0RAOJFU4DfHJ2ew#/registration).
## Sign up for Applied / Comms With AI
The practical guide for communications leaders navigating AI
Subscribe
Email sent! Check your inbox to complete your signup.
No spam. Unsubscribe anytime.
### The Cost of a Miss: Comms With AI in the Govern Phase
URL: https://www.appliedcomms.ai/ai-agent-series-govern/
Last updated: 2026-05-27T06:00:12.000Z
There's no better way to kick off a deep dive than with an uncomfortable arithmetic problem. Here goes...
If a Create-focussed agent helps your team produce three times the content it used to, someone is now reviewing three times the content. The drafting step got an agent. The reviewing step, in most teams, did not. So the volume problem the last article (['The Volume Problem: Comms With AI in the Create Phase'](https://www.appliedcomms.ai/comms-with-ai-in-the-create-phase/)) described does not disappear when a draft is finished. It moves one phase downstream and lands on whoever signs the work off.
That is the Govern phase, and it is the part of the [Comms With AI Operating System](https://www.commswith.ai/os/) teams are least keen to talk about. Strategise is interesting. Create is visible. Govern is the bit that feels like brakes. It is approvals, compliance, fact-checking, tone audits, accessibility, the careful and extremely unglamorous work of making sure nothing goes out that should not have. Few build an AI strategy because they are excited about approval workflows.
**However, Govern is where AI-assisted communications either hold up or quietly fall over.** It is the phase where AI introduces new ways to fail before it offers any way to cope with the extra load. And it is the phase where the argument for using agents is less obvious – and therefore more interesting – than it first looks.
\[If you are joining the series here, the [overview introduces the Operating System and its five phases](https://www.appliedcomms.ai/ai-agent-driven-communications-practical-framework/), [Part 2 goes deep on Strategise](https://www.appliedcomms.ai/ai-agent-series-strategise/), and [Part 3 covers Create](https://www.appliedcomms.ai/ai-agent-series-create/). You can [find the full series here](https://www.appliedcomms.ai/tag/ai-agents-series/). This fourth article is about the control layer that protects everything the first three phases produced.\]
---
## **The Governance Gap**
Communications has always had a governance bottleneck. Approval was regularly slow and painful *long* before anyone had heard of an AI agent. Legal needed to see the claims. Brand needed to see the tone. A senior leader needed to see anything that touched reputation. The work queued, and the queue was the price of not making expensive mistakes in public.
AI did two things to that queue at once, and only one of them was helpful.
Unhelpfully, it multiplied the volume of work arriving at the gate. More drafts, more formats, more variants, all produced faster than the review layer was built to absorb. A governance process calibrated for one press release a week strains badly when it meets a dozen.
Additionally, it has changed the kind of error that can reach the gate. A human writer under deadline pressure makes predictable mistakes. They get a date wrong, they over-claim slightly, they reach for a tired phrase. A communications professional reviewing that work knows what to look for. An AI agent fails differently. It will state a fabricated statistic with total confidence. It will attribute a quote to a real person who never said it. It will cite a source that does not exist, formatted immaculately. It will drift, paragraph by paragraph, toward the bland statistical centre of its training data, so that the tone is not wrong exactly, just no longer anyone's.
These are not failures of effort. They are failures of a different shape, and therefore, a review process designed to catch human error is not set up to reliably catch them. That is the governance gap: production capability scaled, the volume and the failure modes both changed, and the checking layer stayed roughly where it was.
**The hard truth of the AI-in-comms story is that most teams have upgraded Create and left Govern alone. That is the most dangerous possible configuration.** It is faster production feeding an unchanged, now overloaded, gate.
---
## **What the Govern Phase Covers**
Govern is the control layer of the Operating System. It holds everything that stands between a finished draft and a published asset the organisation is willing to defend:
- **Approval workflows**: routing, sequencing, sign-off, status tracking, the coordination of who sees what and in what order
- **Tone and brand-voice verification**: checking that a piece sounds like the organisation, consistently, across volume
- **Claims substantiation**: fact-checking, source verification, making sure every assertion can be stood behind
- **Accessibility and inclusive language**: reading level, alt text, plain language, contrast, jargon
- **Compliance and regulatory review**: sector rules, legal exposure, disclosure requirements, including disclosure of AI use itself
- **Crisis preparation**: pre-approved response frameworks, holding statements, and the governance done in advance so it does not have to be done under fire
The thread running through all of it is that Govern is where reputational risk is priced. Every other phase produces work. Govern decides whether the organisation can live with it.
---
## **The 25/75 Rule for Govern**
Across the CWAI Operating System, the working rule is roughly 30% AI, 70% human. Govern sits below that average, at **25% AI, 75% human**.
This ratio shifts across the five phases:
- **Strategise: 20/80** – research and synthesis AI, judgement human
- **Create: 35/65** – drafting AI, voice and nuance human
- **Govern: 25/75** – systematic checking AI, reputational judgement human
- **Monitor: 40/60** – pattern detection AI, interpretation human
- **Transform: 20/80** – readiness assessment AI, change leadership human
The reason Govern keeps the human share high is not sentiment. It is asymmetry. In Create, a weak agent draft costs you an hour of editing. In Govern, a missed claim costs you a correction, a news cycle, and a quantity of trust that takes far longer to rebuild than any agent ever saved. The maths of governance is lopsided. The downside of a miss dwarfs the upside of a fast pass, so the phase has to be run conservatively by design.
The instinct is that high stakes mean keeping AI out. Govern is reputational; therefore, it's the last place you would let an agent near. This instinct isn't entirely mistaken: it is right that humans must own the reputational calls. However, it is backwards about where governance actually fails.
Governance failures, in practice, are rarely failures of judgement. They are failures of consistency. The claim nobody thought to check. The alt text nobody added. The terminology that drifted across a campaign it was drafted by three people. The disclosure line that was in the template and got cut. These are not hard calls anyone got wrong. They are routine checks a tired human skipped at five o'clock on a Friday, or due to the umpteen other reasons that they were pressed for time.
A notable upside for AI in 2026, is that consistency is the one thing an agent is reliably good at. (Something you would not have said even a year ago.) It does not get tired. It does not skip the boring check on the forty-first asset. It applies the same standard to the all-staff email as to the flagship report. So the 25% an agent does in Govern is not the judgement. It is the floor. It is making sure the routine, checkable, consistency-dependent work actually gets done every single time, which frees the human 75% to spend its attention on the calls that genuinely need a person.
The risk is not putting agents into Govern. The risk is upgrading Create and leaving Govern as it was.
---
## **Where Agents Change Governance Work**
Four applications are worth examining in detail, because they are where the change in the Govern phase is most tangible.
### **1\. Claims substantiation and fact-checking**
This is the application that matters most, because it addresses AI's most dangerous failure mode directly.
A claims-substantiation agent does something a human reviewer rarely has time to do properly. It reads a draft and extracts every factual assertion in it, one by one: every number, every named source, every attributed quote, every "studies show", every comparative claim. For each one, it asks a blunt question. Where did this come from, and would a sceptic accept the answer? Claims with a solid source are passed. Claims with a weak source are flagged. Claims with no source, or a source the agent cannot verify, are escalated for a human to resolve before the piece moves.
The principle underneath this is worth stating plainly, because it is the single most important governance design decision in AI-assisted comms. **The agent that drafts a piece must not be the only agent that checks it.** If the same system, with the same prompt and the same blind spots, both writes and verifies, its errors are correlated. It will confidently wave through its own fabrications. Effective claims governance uses a separate checking layer, ideally a different model with a different instruction set, whose entire job is to be the sceptic. Separation of duties is an old governance idea. It applies cleanly here.
This is also where the [interview with NOAN's Neal Mann](https://www.appliedcomms.ai/noan-neal-mann-interview/) on this site is worth revisiting. His argument, that most AI-powered communications is built on sand because it has no verified fact layer underneath it, is precisely the problem a substantiation agent exists to address. The agent does not create the fact layer. It is the discipline that refuses to let a piece proceed without one.
The [Claims Substantiation Checklist on CommsWith.AI](https://www.commswith.ai/library/governance/claims-substantiation-checklist/) is built to structure this work, so the agent's output is a usable list of resolved and unresolved claims rather than a vague reassurance.
### **2\. Tone and brand-voice auditing**
Part 3 described how a voice-trained agent can help produce content in a specific spokesperson's or organisation's voice. The same voice profile has a second job, and it belongs in Govern.
A brand-voice audit agent takes a finished draft and checks it against a defined voice profile: the organisation's characteristic register, the words it uses and the words it avoids, the level of formality, the rhythm. It flags drift. It catches the paragraph that has slid into corporate neutral, the sentence that is technically on-brand but tonally off for this particular moment, the piece that reads as though it could belong to any organisation in the sector.
This matters more as volume rises. When a team published one considered piece a week, voice consistency was held by the simple fact that one or two people touched everything. When a team publishes at agent-assisted volume, no single person reads all of it, and voice consistency stops being automatic. The audit agent is how you hold a standard you can no longer hold by hand.
It is worth saying that this is not a new idea on Applied Comms AI. The ["Will My Boss Hate This?" stakeholder review tool](https://www.appliedcomms.ai/ai-stakeholder-comms-review-tool/) built earlier on this site was, in effect, an early governance agent: a structured check that ran a draft against the predictable reactions of the people who would have to approve it. That is Govern work. The [Tone and Style Checker](https://www.commswith.ai/library/governance/tone-style-checker/) and [Brand Voice Audit Checklist](https://www.commswith.ai/library/governance/brand-voice-audit-checklist/) templates carry the same logic into a repeatable form.
### **3\. Accessibility and inclusive-language scanning**
This is the lowest-risk, highest-consistency win in the entire phase, and it is routinely skipped.
Accessibility checking is exactly the kind of work humans do badly and agents do well. It is systematic. It is rule-based. It is tedious. It is the same set of checks on every asset: is the reading level appropriate for the audience, is there meaningful alt text on every image, is the language plain where plain language is needed, is jargon defined, does the structure work for a screen reader, is the inclusive-language standard met. A human reviewer under deadline pressure will do this properly on the important pieces and quietly let it slide on the rest. An agent does it on all of them, closer to identically, in seconds.
### **4\. Approval workflow orchestration**
The fourth application is the least discussed and, for many teams, the one that removes the most friction.
Most of the pain in approvals is not the reviewing. It is the coordination around the reviewing. Who needs to see this. In what order. What does each reviewer actually need: legal wants the claims and the contracts, brand wants the tone and the visual treatment, the executive sponsor wants the strategic risk and does not want a copy-edit. Chasing. Reminding. Version control. Knowing what is stuck and why.
An orchestration agent handles that coordination layer. Given a piece and a workflow definition, it assembles the right review pack for each reviewer, so legal is not wading through brand notes, routes the piece in the correct sequence, tracks status, and flags what is overdue. It does not approve anything. It removes the administrative drag that makes approval feel slow even when the actual reviewing is fast.
The [Approval Workflow Mapper](https://www.commswith.ai/library/governance/approval-workflow-mapper/) and [Content Approval Tracker](https://www.commswith.ai/library/governance/content-approval-tracker/) templates are designed to define the workflow first, in plain terms, so that what the agent orchestrates is your actual governance process and not an invented one. Map before you automate. An automated bad workflow is just a bad workflow that now runs faster.
---
## **A Test Your Governance Should Pass**
Before a piece is published, run it against the following five questions. These are the questions that decide whether a governance process is real or decorative.
**1\. For every type of risk in this piece, can you name who is accountable?**
Factual risk, legal risk, brand risk, reputational risk. If the answer is "the team reviewed it", no one is accountable. Governance without named ownership is a group of people each assuming someone else caught it.
**2\. Is every factual claim traceable to a source a sceptic would accept?**
Not a source you trust. A source someone who wants you to be wrong would still have to concede. If a claim cannot meet that bar, it is an opinion, and it should be written as one or removed.
**3\. Would this piece survive being read aloud by your most hostile stakeholder?**
Pick the specific person: the journalist with a grudge, the regulator, the board sceptic. Read it as they would. If a sentence hands them a weapon, you want to know now, not after publication.
**4\. Is your use of AI disclosed where disclosure matters?**
This series discloses its own AI use deliberately. Your organisation needs a clear position on when AI assistance is disclosed, to whom, and in what form. "We never decided" is not a position. It is an exposure.
**5\. If this went wrong in public, is your response already drafted?**
For routine content the honest answer is that it will not go wrong, and that is fine. For anything sensitive, anything making a strong claim, anything on a contested topic, the response should exist before the piece does. If it does not, the piece is not finished.
If the work passes all five, it is ready to publish. If it fails any of them, the fix belongs in Govern, before the piece goes out, where the cost of closing the gap is measured in minutes rather than news cycles.
---
## **Where To Start**
If Govern is the phase your team has under-invested in, which it very likely is, because production is where the visible excitement sits, three starting points will pay back quickly.
**Start with accessibility and claims checking.** These are the lowest-risk, highest-consistency agent applications in the whole Operating System. The work is rule-based, the agent does it identically every time, and the failure you are preventing, an unsupported claim or an inaccessible asset, is both common and avoidable. If you do one thing in Govern, make it this.
**Build one brand-voice profile and run it as a standing check.** Not a document describing your tone in adjectives. A working profile, built from real pieces that sound right, used by an audit agent on outgoing content. It is the only practical way to hold voice once you are publishing at agent-assisted volume.
**Map your approval workflow before you automate any of it.** Write down who reviews what, in what order, and why, in plain language. Most teams discover their real workflow is not the one they think they have. Fix the workflow first. Automate it second. An orchestration agent should accelerate a good process, never entrench a bad one.
The templates in the Govern phase on CommsWith.AI are built around this logic. Each sits at the point where systematic agent checking meets human reputational judgement, with the split held where it belongs.
[**Browse the full Govern phase on CommsWith.AI**](https://www.commswith.ai/library/governance/)**.**
---
## **About**
[Applied Comms AI](https://www.appliedcomms.ai/) is the practical guide for communications leaders navigating AI, grounded in hands-on experimentation, workflow transformation and real-world implementation. Read the full [AI Agent series here](https://www.appliedcomms.ai/tag/ai-agents-series/). [CommsWith.AI](https://www.commswith.ai/) is the companion template and resource library for communications professionals using AI, and the [Govern phase library](https://www.commswith.ai/library/governance/) covers approval workflows, tone and brand-voice verification, claims substantiation, accessibility and crisis preparation. Both sit alongside [Faur](https://faur.site/), a communications consultancy pioneering practical AI expertise for organisations ready to implement at scale. If your team is working on the Govern phase and needs bespoke support, whether governance design, claims and compliance workflows or crisis-readiness work, get in touch at michael@faur.site or connect with me on [LinkedIn](https://www.linkedin.com/in/michaelmaclennan).
---
*This article is part of the* [*AI Agent series published on Applied Comms AI*](https://www.appliedcomms.ai/tag/ai-agents-series/)*. The series maps to the* [*Comms With AI Operating System*](https://www.commswith.ai/start-here/)*: Strategise, Create, Govern, Monitor, Transform.*
### The Volume Problem: Comms With AI in the Create Phase
URL: https://www.appliedcomms.ai/comms-with-ai-in-the-create-phase/
Last updated: 2026-05-05T07:07:03.000Z
Most conversations about AI in communications live in the Create phase. This is the phase people think of when they think of AI in comms: the press release draft, the LinkedIn post, the blog article, the executive comment, the internal memo. It is the obvious territory, and it is where most teams have already started to experiment.
It is also where the honest conversation about agents is hardest to have.
The temptation with Create is to measure the value of AI by speed. First draft in minutes. Social pack in an afternoon. Campaign assets in a day. Those numbers are real and they matter – [I’ve shown what can be done at pace previously](https://www.appliedcomms.ai/claude-skills-projects-business-pitch/) – but they describe capacity, not quality. The harder question is whether the work coming out of an agent-assisted production line holds up against the standard senior communicators used to apply to work that took several times longer. Sometimes it does. Often it does not. **Knowing the difference is the skill that defines whether a team gets lift from Create-phase agents or drifts into polished mediocrity at scale.**
If you missed the earlier articles in this series, [the overview introduces the Operating System and its five phases](https://www.appliedcomms.ai/ai-agent-driven-communications-practical-framework/), and [Part 2 goes deep on Strategise.](https://www.appliedcomms.ai/ai-agent-series-strategise/) This third article is about the phase where most teams meet AI for the first time, and where the judgement calls are most consequential.
---
## **The Shape of the Create Phase**
Create covers everything that happens between a signed-off strategy and a published asset:
- **First-draft production:** press releases, bylines, LinkedIn posts, speeches, executive comments, internal announcements, customer communications, statements, and briefings
- **Multi-format packaging:** translating one piece of thinking into the six to twelve formats a modern campaign needs, with appropriate adjustments for audience, channel, length, and register
- **Editorial planning:** content calendars, thought leadership plans, campaign timelines, and channel sequencing
- **Narrative development:** building the connective tissue between individual pieces so a campaign reads as one coherent argument rather than a shopping list of outputs
- **Ghostwriting and voice work:** producing content in the voice of a specific spokesperson or organisational persona
The thing that unites all of this is that the work is visible. Every Create-phase output has an audience. When quality slips here, it slips in public. That is why the balance between agent capability and human judgement is different in Create from every other phase of the Operating System.
---
## **The 35/65 Rule for Create**
Across the Operating System, the working rule is roughly 30% AI, 70% human. Create shifts the balance slightly toward AI: **35% AI, 65% human**.
Agents do more of the work in Create than they do in Strategise or Govern because the task shape suits them. First drafts from a defined brief. Multi-format adaptation from a strong master. Structural scaffolding where tone is already set. These are tasks with clear inputs and outputs, where consistency matters, and where volume is a meaningful constraint.
What stays human is the work that makes content worth publishing. Voice. Judgement on nuance. The decision to cut a line that sounds clever but lands wrong. The sense of which sentence is pulling too much weight and needs to be broken into two. The instinct that says this draft is fine but not quite there yet, and the craft to know what to do about it.
The 35/65 split is not about time. It is about the share of the work that previously sat with a human. The human share does not get smaller; it gets more concentrated on the decisions where experience is load-bearing.
---
## **Where Agents Change Creative Work**
Four applications are worth looking at in detail, because they are where the change in the Create phase is most tangible.
### **1\. Structured first drafts from a strong brief**
The difference between a useful agent-generated draft and a useless one is almost always upstream. If the message house is clear, the audience is defined, the call to action is agreed, and the length is set, an agent can produce a first draft that a senior writer can take to final in under an hour. If any of those inputs is vague, the draft will read as if it was written by a committee of none of the right people.
This is why Create-phase agent work is inseparable from Strategise-phase rigour. The teams getting the strongest lift from first-draft agents are the teams that have invested in their upstream work. The teams getting weak results are almost always compensating for thin strategy by asking the agent to fill the gap. It cannot.
The [Press Release Structure on CommsWith.AI](https://www.commswith.ai/library/content/press-release-structure/) is built for this workflow: making the upstream inputs explicit, so the agent has something substantive to work with.
**Tools that work well here:** Claude for nuance and structural reasoning; ChatGPT for speed on familiar formats; Gemini where the work needs to be tied to live data or search results. In my own testing, Claude is the most consistent for drafting that needs to hold a specific voice across several pieces in a row.
### **2\. Multi-format packaging from a master piece**
Most communications teams spend a disproportionate amount of time turning one idea into many outputs. A single keynote becomes a press release, three LinkedIn posts, a blog article, an email to members, a talking-points document, and a partner briefing. Each needs its own tone, length, and structure. Each takes time.
Agents are excellent at this task. Given a strong master piece and a format brief, a packaging agent can produce a coherent multi-format set in minutes. The output is rarely final, but it is close enough that the editing time is a fraction of the drafting time. For teams running frequent campaigns, this is the single highest-velocity use of Create-phase agents.
The practical catch: the master piece has to hold up on its own. If it hedges, the derivatives hedge harder. If it buries the point, every format buries the point. Multi-format packaging is a magnifier, not a fixer. Once you’ve done this, produce a checklist for the criteria all outputs must individually meet so that they are fully tailored and optimised for the platform of choice.
### **3\. Voice work and ghostwriting**
The least obvious but most interesting application. Senior executives are increasingly expected to have a LinkedIn presence, a speaking profile, and a visible public voice, at a frequency that no real human can sustain without help. Communications teams have always ghostwritten for leaders. Agents are starting to change what that ghostwriting looks like.
A voice-trained agent, given a corpus of a spokesperson’s previous writing, can produce drafts that carry their sentence rhythms, preferred framings, and characteristic phrases. The quality varies enormously. A light prompt gives you a generic executive voice with the spokesperson’s name on it. A deeply worked voice profile, built from a dozen representative pieces with notes on what makes them sound like the person, gets you something that stands up to scrutiny.
The judgement call here is ethical before it is technical. Ghostwriting has always involved a degree of fiction about authorship; agents make that fiction easier to scale and harder to police. Teams using voice-trained agents need a clear internal position on disclosure, review, and when a piece should not be produced at all.
### **4\. Editorial planning and narrative architecture**
The fourth application is the most senior and the least discussed. Agents are useful for building editorial plans that hang together as a narrative, rather than as a content calendar with dates attached. Given a six-month positioning objective and a list of input events (earnings days, conferences, product moments, regulatory windows), an agent can produce a sequenced editorial plan that ladders individual pieces back to the strategic goal.
This is closer to strategic work than to drafting, and the balance tilts back toward 25/75 here. The agent’s value is in handling the structural complexity of sequencing a dozen moving parts. The human value is in judging which pieces carry the argument forward and which are filler.
The [Monthly Content Calendar Planner on CommsWith.AI](https://www.commswith.ai/library/content/content-calendar-monthly/) is designed to structure this work so the agent’s output is a draft plan rather than a list of dates.
---
## **Three Examples From Practice**
The following are drawn from real client work, anonymised and simplified.
### **Example 1: A scale-up scaling founder voice**
A Series B software company wanted to triple the volume of its founder’s public writing over six months, to establish her as a credible voice on AI governance in the regulated sectors the company sells into. The comms lead had a two-person team and was already at capacity.
The Create-phase work used a voice-trained agent, built from twelve pieces of the founder’s previous writing and two hours of interview transcript. The agent produced first drafts of LinkedIn posts, byline articles, and speaking abstracts; the comms lead edited them with the founder in thirty-minute sessions rather than the two-hour drafting sessions that had previously been the pattern.
Over six months, published output rose from roughly one piece per fortnight to three per week, across LinkedIn, trade press, and speaking platforms. Share of voice in the founder’s target topic rose measurably; inbound speaking invitations tripled. The founder’s stated view at the six-month review was that the published work still sounded like her, which was the non-negotiable test.
### **Example 2: A charity responding to a contested news cycle**
A national charity found itself unexpectedly at the centre of a fast-moving story about a policy change. Over a seventy-two-hour period, it needed to produce a holding statement, a full position paper, six social variants for different platforms, an all-staff briefing, a partner-network briefing, and three spokesperson Q&A packs for different media tiers.
The Create-phase workload in a traditional setup would have required the team to work twenty-hour days or to cut one of the deliverables. Instead, a packaging agent produced first drafts of every format from an agreed position paper master, within four hours. The team spent the remaining sixty-eight hours on strategic refinement, spokesperson preparation, and the judgement-heavy work of deciding what to say and to whom.
Every published piece was substantially edited by a human before going out. Nothing was published in the first draft the agent produced. But the drafts saved roughly a day and a half of writing time, which was the difference between responding well and responding late.
### **Example 3: A professional services firm producing a flagship report**
A mid-sized professional services firm produced an annual flagship report. The report had historically taken six months, involved twelve internal contributors, and had a production pattern of scope creep, late copy, and last-minute edits that pushed the launch date back two weeks every year.
The Create phase was restructured around three agent applications. A structuring agent produced a detailed outline and chapter scaffolds from the agreed brief before any contributor wrote a word. A drafting agent took each contributor’s raw input and produced a first pass that matched the report’s editorial voice. A packaging agent produced the derivative outputs (executive summary, press release, LinkedIn series, speaking abstracts) from the final report, in parallel with the report’s own production.
The report shipped on time. Internal contributors reported significantly less frustration with the process, because the agent handled the editorial heavy lift that had previously been done by a thinly stretched in-house writer. The quality of the final report was judged by the managing partner to be comparable to previous years; the efficiency gain was roughly five weeks of production time.
The lesson from this example is not that agents replaced the writer. The writer’s role shifted. Less time on initial drafting, more time on structural editing and voice consistency. The quality floor held, and the quality ceiling held.
---
## **A Test Your Create Work Should Pass** 🧪
Before a piece goes into the Govern phase, run it against the following five questions. These are the questions a senior editor at a national newspaper would ask, stripped of politeness. If the piece fails any of them, it is not ready.
### **1\. Could a reader explain what you are arguing after one pass?**
Not after a second read. Not after re-reading the headline. After one pass. If the argument only surfaces on a re-read, the piece is structurally weak. Make the first third do more work.
### **2\. Is there a specific proof point in the first third that stops a sceptic from skimming?**
Not a claim. A proof point. A number, a name, a specific example, a verifiable outcome. If the first third is throat-clearing, the piece will not hold the reader long enough to make its case.
### **3\. Does the piece sound like a human you would recognise?**
Read it aloud. If three paragraphs in a row could have been written by anyone, the voice has collapsed. Agents are particularly prone to this, because they default to the statistical centre of their training data. Every paragraph needs at least one move that sounds like the author.
### **4\. Is there one line that would survive being quoted back at you in six months?**
The best communications work has a line that outlives the campaign. If nothing in the piece is quotable, the piece is probably forgettable. If everything is trying to be quotable, the piece is probably exhausting.
### **5\. What would you cut if you had to lose ten per cent?**
Every piece has ten per cent of weight it could lose. If you cannot identify what it is, the piece is over-written and the reader will feel it. If cutting it changes nothing meaningful, cut it.
If the piece passes all five, it is ready to move into Govern. If it fails any of them, the fix belongs in Create, not downstream.
---
## **Where To Start**
If the Create phase is where your team feels the most volume pressure, three starting points will pay back quickly.
**Start with multi-format packaging.** It is the clearest demonstration of agent capability, the lowest-risk application, and the fastest time to visible lift. If your team is spending more than a day packaging a single announcement into its full channel set, a packaging agent will pay for itself in the first week.
**Build one voice profile, properly.** Pick your most active spokesperson. Invest a full day in building a voice profile that holds up. Use it on three pieces, iterate, and extend to a second spokesperson only when the first is working. Voice work done poorly is worse than no voice work at all; done properly, it is one of the highest-leverage Create investments.
**Resist the volume trap.** The temptation with Create-phase agents is to produce more. The better move is often to produce the same amount and raise the quality floor. Teams that use agents to ship more mediocre content are measurably worse off than teams that use agents to ship the same volume at a higher standard. The metric that matters is not output per week; it is how often a senior leader says the work read like it belonged to the brand.
The templates in the Create phase on CommsWith.AI are built around this logic. Each sits at the intersection of agent capability and human judgement, with prompts and review checklists designed to keep the split in the right place.
[**Browse the full Create phase on CommsWith.AI**](https://www.commswith.ai/library/content/)**.**
---
## **What Comes Next**
Article 4 in this series covers the **Govern** phase: risk, quality, compliance, and the approval workflows that keep the Create phase from becoming a liability. It is where the 25/75 split kicks in, and where the argument for agent-supported governance is less obvious than it looks.
If your team is working on Create-phase implementation and you want to compare approaches, I am happy to hear from fellow practitioners at [michael@faur.site](mailto:michael@faur.site).
---
## **About Applied Comms AI**
[Applied Comms AI](https://www.appliedcomms.ai/) is the practical guide for communications leaders navigating AI, grounded in hands-on experimentation, workflow transformation, and real-world implementation. Read the full [AI Agent series here](https://www.appliedcomms.ai/tag/ai-agents-series/).
---
## **About Comms With AI**
[CommsWith.AI](https://www.commswith.ai/) is the template and resource library for communications professionals using AI. The [Create phase library](https://www.commswith.ai/library/) covers first-draft production, multi-format packaging, voice work, and editorial planning.
---
## **About Faur**
[Faur](https://faur.site/) is a communications consultancy pioneering practical AI expertise for organisations ready to implement at scale. If your team is working on the Create phase and needs bespoke support, voice profile development, workflow redesign, or agent deployment, [get in touch](applewebdata://1FC8B5FA-A89A-4E6B-BFF1-514380245E71/info@faur.site).
---
*This article is part of the* [*AI Agent series published on Applied Comms AI*](https://www.appliedcomms.ai/tag/ai-agents-series/)*. The series maps to the* [*Comms With AI Operating System*](https://www.commswith.ai/start-here/)*: Strategise, Create, Govern, Monitor, Transform.*
### The Work Before the Work: Comms With AI in the Strategise Phase
URL: https://www.appliedcomms.ai/ai-agent-series-strategise/
Last updated: 2026-09-07T10:24:01.000Z
Most AI conversations in communications start in the wrong place – like you're tuning in only right at the end of your favourite soap opera.
They start with the draft – the press release, the social post, the executive briefing – and ask how AI can make the drafting faster. That framing concedes too much. It assumes the quality of the output depends mainly on how well the writer handles the last mile.
**In practice, the quality of any communications work is set long before the drafting starts. It is set in the research, the positioning, the stakeholder mapping, the message architecture, the objectives.**
It is set in the work that senior communicators do when they are thinking rather than typing. When that upstream work is weak, no amount of polish on the draft can save it. When it is strong, even a rough draft tends to land.
This is the Strategise phase of the [Comms With AI (CWAI) Operating System](https://www.commswith.ai/start-here/): the intelligence and planning layer that informs everything downstream. It is where AI agents are starting to change how senior leaders think, not just what they produce.
If you missed the first article in this series, [it introduced the Operating System and its five phases](https://www.appliedcomms.ai/ai-agent-driven-communications-practical-framework/): Strategise, Create, Govern, Monitor, Transform. You can [find the full series here](https://www.appliedcomms.ai/tag/ai-agents-series/). This second article goes deep on the first phase — where most strategic communications work actually begins.
---
## The Strategic Rigour Problem
Senior communicators have always lived with a quiet tension.
The strategic groundwork they know the work needs – proper audience analysis, rigorous stakeholder mapping, a defensible message architecture, a clear landscape read – takes time. The brief they have been handed, or the crisis they are responding to, or the launch they are supporting, often does not allow for that time.
What gets compressed first is the thinking. The kick-off meeting becomes the strategy. The gut feel becomes the audience insight. The creative director's instinct becomes the positioning. The work ships on time, and owing to experience and savvy know-how will still be of high quality, but the rigour behind it is thinner than anyone would admit in public.
This is not a failure of the team, far from it! It is a structural constraint. Strategic work at the level it deserves has never been scalable, because it depends on experience, synthesis, and judgement. Until recently, there was no way to do it faster without doing it worse.
AI Agents start to change that equation. Not by replacing the judgement – that part remains resolutely and stubbornly human – but by collapsing the time spent on the groundwork that informs it.
---
## What the Strategise Phase Covers
Strategise is the broadest phase of our CWAI Operating System. It holds everything that happens before a single sentence gets drafted:
- **Research and intelligence**: landscape analysis, competitive monitoring, trend identification, sectoral reading
- **Audience work**: segmentation, profiling, psychographic analysis, channel behaviour mapping
- **Stakeholder engagement**: mapping, prioritisation, engagement planning, influence analysis
- **Positioning and messaging**: message architecture, narrative design, proof point organisation, differentiation
- **Planning**: campaign briefs, communications plans, channel strategy, objectives, measurement frameworks
On [CommsWith.AI,](https://www.commswith.ai/) the Strategise phase currently holds 20 templates — the largest of any phase. That density reflects the reality of senior communications work: most of the intellectual load sits upstream of production.
Messaging frameworks, audience profiles, and stakeholder maps are strategic outputs, not creative ones – it's also all an indicator of how central communications is to well-performing companies. (All us comms pros can and should take a bow here, though not for too long – after all, time is pressing...)
---
## The 20/80 Rule for Strategise
Across the CWAI Operating System, the general rule is roughly **30% AI, 70% human judgement**. Strategy tilts further toward the human side. For the Strategise phase specifically, the right (still rough) balance is closer to **20/80**. AI handles research, structuring, and first-pass synthesis; humans handle interpretation, prioritisation, and judgement about what matters.
This ratio shifts across the five phases:
- **Strategise: 20/80** – research and synthesis AI, judgement human
- **Create: 35/65** – drafting AI, voice and nuance human
- **Govern: 25/75** – systematic checking AI, reputational judgement human
- **Monitor: 40/60** – pattern detection AI, interpretation human
- **Transform: 20/80** – readiness assessment AI, change leadership human
Strategy is where AI earns its place least automatically. Agents can give you a landscape map in an hour that would take a researcher two days. They cannot tell you which parts of that map matter, which stakeholders to prioritise, or which proof point will land with a sceptical board. That is the 80%. Getting the 20% right still buys you meaningful capacity.
---
## Where Agents Change Strategic Work
Four specific applications are worth examining in detail, because they are the areas where the change is most tangible.
### 1\. Landscape analysis and competitive intelligence
Traditional landscape work involves a researcher trawling through competitor websites, annual reports, media coverage, and analyst commentary to build a view of how a market is positioning. It is thorough, necessary, and slow.
A landscape analysis agent can run the same scan in a fraction of the time. Give it five competitors and a focus area (positioning, narrative pillars, channel mix, tone of voice), and it will return a structured comparison in under an hour. The [Competitor Comms Audit template on CommsWith.AI](https://www.commswith.ai/library/strategy/competitor-comms-audit/) is built for this kind of agent-supported analysis — it defines the structure so the agent's output is usable immediately.
The catch: agents surface what is publicly available. They do not know which competitor is about to pivot, which is struggling internally, or which is about to lose a key executive. That context still comes from network and experience.
**Tools that work well here:** Perplexity for synthesised web research, Claude or ChatGPT with deep research modes for structured competitive analysis, Gemini for Google-adjacent intelligence. Each has strengths: in my own testing, Perplexity is fastest for surface scans, ChatGPT is best for structured analysis output, and Gemini is strongest when the analysis involves Google Search data specifically.
### 2\. Stakeholder mapping at scale
Stakeholder maps are one of the most undervalued strategic tools in communications. Done properly, they are the foundation for every engagement decision downstream: who gets what message, in what order, through what channel, with what rationale.
Done quickly, they are a list of names on a power/interest grid (which usually then sits hidden away in a shared folder, unloved and unused).
Agents sit between the two. They can take a list of stakeholders, pull public information about each, and produce a structured assessment across dimensions that would take a junior strategist a full day to compile manually: current public position, channel preferences, recent statements, likely concerns, probable response to the organisation's planned message.
The [Stakeholder Mapping Matrix on CommsWith.AI](https://www.commswith.ai/library/strategy/stakeholder-mapping-matrix/) provides the structure; agents do the legwork. The strategic call – who actually matters, who to engage first, where the risks are – remains with the team.
Do bear in mind that agents are still prone to getting details wrong – and occasionally conjuring up details (and in extreme cases even imaginary people). You still want/need that human review to vet details, and also add anything else that has been missed. This remains a time-saving tool, definitely not a one-and-done solution.
### 3\. Audience profiling from unstructured data
Audience work used to be a specialist discipline requiring research budgets and timelines that most comms teams do not have. The shift to agent-assisted profiling has been one of the more practical wins of the last twelve months.
Feed an agent a collection of unstructured inputs – survey responses, sales call transcripts, customer support tickets, social listening data, reviews – and ask it to build audience segments with consistent structure: motivations, barriers, language patterns, channel preferences, trigger moments.
The output is not a replacement for proper qualitative research, but it is a dramatic step up from the assumptions that usually stand in for audience insight in fast-moving work. It also throws up insights you may not have previously considered, and of course you can then converse with your tools of choice to dig deeper.
The [Comms Audience Profile template](https://www.commswith.ai/library/strategy/comms-audience-profile/) and [Audience Segmentation Worksheet](https://www.commswith.ai/library/strategy/audience-segmentation-worksheet/) on CommsWith.AI are both designed to take agent-generated profiles and structure them for team use.
### 4\. Message architecture and positioning
Positioning work is where many senior communicators are most sceptical about AI, and with good reason. Positioning involves trade-offs, long-horizon judgement, and sensitivity to organisational politics that agents cannot touch.
But positioning also involves a lot of structured groundwork: clarifying the competitive set, surfacing differentiation candidates, stress-testing proof points, and generating multiple framings for the same underlying idea. That groundwork is where agents can earn their keep.
The [Message House](https://www.commswith.ai/library/strategy/message-house/), [Positioning Statement Generator](https://www.commswith.ai/library/strategy/positioning-statement-generator/), and [Proof Points Bank](https://www.commswith.ai/library/strategy/proof-points-bank/) templates are built around this workflow: agent produces structured first-pass outputs, human makes the judgement calls about which framing wins, which proof points carry weight, and where the narrative lives.
---
## Three Examples From Practice
The following are drawn from real client work, anonymised and simplified to protect confidentiality. They show how the Strategise phase operates in practice, across different sectors and timelines.
### Example 1: A membership body navigating a regulatory shift
A professional membership body commissioned strategic communications support ahead of an anticipated regulatory change that would affect around 60% of its members' working practices. The team had six weeks to build a position, develop a narrative, and prepare member-facing communications before the change was announced.
The Strategise phase ran in the first ten days. Agent-supported landscape analysis mapped how comparable bodies in adjacent sectors had handled previous regulatory shifts — which positions had held up well, which had backfired, and which had quietly been abandoned. A parallel stakeholder mapping exercise identified thirty named individuals across regulators, media, member associations, and vocal members whose early response would shape perception.
The agent did the compilation. The strategic call — to lead with a specific framing rather than the obvious one — was made in a 90-minute working session with the CEO and chair, drawing on the agent-produced groundwork but departing from its implied direction on two key points. The work that may have taken a research team two weeks – a window way too wide for this situation – was compressed into a couple of days. The weeks that followed could then be focussed on areas including narrative refinement, spokesperson preparation, and member engagement.
### Example 2: A scale-up preparing for a funding announcement
A Series B software company needed to position an eight-figure funding round. The comms lead had two weeks, limited research support, and a founder who wanted the announcement to land in tier-one business media as well as sector press.
The Strategise phase began with agent-assisted competitive intelligence on how comparable companies had positioned their recent rounds: which narratives had landed, which had been ignored, and which spokespeople had been treated as credible in business media versus sector media. The agent also profiled the journalists most likely to cover the story, drawing on their recent output to identify which angles would resonate and which would not.
The resulting message architecture distinguished between the sector-media narrative (product capability and customer outcomes) and the business-media narrative (market opportunity and founder story). The founder initially pushed for a single unified message; the agent-produced evidence that the two audiences responded to different framings carried more weight in the strategic conversation than the comms lead's instinct alone could have.
The announcement landed in three tier-one outlets and six sector publications, with distinct but aligned framings in each. The strategic differentiation was the single highest-impact decision in the campaign; the agent work supporting it took six hours over two days.
### Example 3: A non-profit planning a three-year campaign
A national charity commissioned a three-year awareness and behaviour-change campaign on a contested social issue. The Strategise phase ran for six weeks before any production work began: unusually long, because the strategic stakes were high and the topic's sensitivity meant mistakes would be expensive.
Three pieces of agent-supported work shaped the campaign. First, a landscape analysis of every major campaign on the same or adjacent issues over the previous ten years: what had moved public attitudes, what had backfired, and what had plateaued. Second, a stakeholder mapping exercise covering 120+ named individuals and organisations across policy, media, academia, and lived-experience communities. Third, a message architecture exercise that tested 15 framings against the audience insight, narrowing to three that the team then refined manually.
The agent work did not make the strategic decisions. It compressed the time spent on the inputs to those decisions from what could have been months to six weeks, and raised the quality floor of the analysis, making the team's strategic conversations better informed. The campaign is now in year two, tracking ahead of its attitudinal targets, though outcome metrics will not be available for another eighteen months.
---
## A Test Your Strategy Should Pass
If you are unsure whether your Strategise work is strong enough, run it against the following five questions before moving to Create. These are not academic; they are the questions a sceptical board member, a sharp journalist, or a new CEO will likely ask within minutes of reading your plan.
**1\. Can you name your primary audience in one sentence, without hedging?**
If the answer contains "and" more than once, you have very likely not segmented properly. "Mid-market CFOs in regulated industries" is a primary audience. "CFOs, COOs, and risk leaders in finance, healthcare, and energy" risks feeling like a spreadsheet.
**2\. What is the single sentence you want your primary audience to repeat after reading your work?**
If you cannot write it down, your team will struggle to deliver it.
**3\. What are the three proof points a sceptic would need before believing your main claim?**
Not your three favourite proof points. The three that would move someone who begins the conversation doubting you, and leaves feeling persuaded in the right direction (if not fully converted!).
**4\. Who are the five stakeholders most likely to shape how this lands, and what do you know about their recent public positions?**
Not who you would like to influence. Who actually influences the outcome.
**5\. How will you know in 90 days whether this worked?**
The objective needs to be specific enough that you can bet on it.
If your strategy passes all five, you are ready for Create. If it fails any of them, the work to fix it belongs in Strategise — not downstream, where the cost of the gap increases the further you go.
---
## Where To Start
If the Strategise phase is the part of the Operating System your team is weakest in – which it likely is, because strategic rigour is where time pressure bites first — three starting points are worth considering.
- **Start with landscape analysis.** It is the lowest-risk agent application, the highest-velocity time saving, and the easiest to evaluate. If your team is spending more than a day on competitive intelligence for any single engagement, agent support will likely pay for itself in the first use.
- **Then move to stakeholder mapping.** The structural nature of the work suits agent support, the templates on [CommsWith.AI](https://www.commswith.ai/) give you a usable framework immediately, and the quality uplift is visible to anyone reviewing the output.
- **Audience profiling and message architecture are the more advanced applications**. They're worth building toward once the simpler workflows are embedded, because they require clearer judgement about when to trust agent output and when to push back on it.
The twenty templates in the Strategise phase on CommsWith.AI are all designed to support this kind of agent-assisted strategic work. They provide the structure that makes agent output usable, and they hold the team's judgement at the points where judgement matters most.
[Browse the full Strategise phase on CommsWith.AI](https://www.commswith.ai/library/strategy/).
---
## Join the Discussion Session
This article has a companion session on Wednesday 29 April at 1.30pm: a free, small-group discussion for senior communicators working on AI implementation in their teams. Not a masterclass — a conversation. A chance to compare notes on how the Strategise phase is playing out in real organisations, surface the practical challenges that do not make it into case studies, and pressure-test how the Operating System applies to your own context.
[Register for the Strategise discussion session on Eventbrite](https://www.eventbrite.com/cc/ai-agents-for-comms-leaders-a-5-part-masterclass-4820359).
Sessions for Create, Govern, Monitor, and Transform will follow the same format as the series progresses.
---
## What Comes Next
Article 3 in this series covers the **Create** phase — where strategy becomes content, at scale and under pressure. It is where the AI contribution rises to 35/65, and where the volume pressures on communications teams are most visible.
If your team is working on the Strategise phase and you want to compare approaches, I am always happy to hear from fellow practitioners at michael@faur.site.
---
## About Applied Comms AI
[Applied Comms AI](https://www.appliedcomms.ai/) is the practical guide for communications leaders navigating AI — grounded in hands-on experimentation, workflow transformation, and real-world implementation. Read the full [AI Agent series here](https://www.appliedcomms.ai/tag/ai-agents-series/).
## About Comms With AI
[CommsWith.AI](https://www.commswith.ai) is the template and resource library for communications professionals using AI. The [Strategise phase library](https://www.commswith.ai/library/strategy/) contains 20 templates covering audience analysis, stakeholder mapping, positioning, message architecture, and campaign planning.
## About Faur
[Faur](https://faur.site) is a communications consultancy pioneering practical AI expertise for organisations ready to implement at scale. If your team is working on the Strategise phase and needs bespoke support — strategic frameworks, agent deployment, or capability development — get in touch at michael@faur.site.
---
*This article is part of the AI Agent series published on Applied Comms AI. The series maps to the Comms With AI Operating System: Strategise, Create, Govern, Monitor, Transform.*
### AI Agent-Driven Communications: A Practical Framework for Transforming How Comms Teams Actually Work
URL: https://www.appliedcomms.ai/ai-agent-driven-communications-practical-framework/
Last updated: 2026-04-22T16:58:18.000Z
Communications has a workflow problem.
Communications teams sit at the convergence of reputation, narrative, stakeholder expectation, and organisational change – right at the heart of what defines the best and strongest organisations.
Meanwhile, the external environment has accelerated – 24-hour news cycles, heightened public scrutiny, digital complexity at scale – but internal workflows have largely not kept pace. Most teams still operate on processes designed for the early social media era. Manual monitoring. Ad hoc content production. Reactive decision-making. Planning cycles that assume more stability than the world currently offers.
AI is often introduced tentatively and tactically: someone asks ChatGPT for a first draft; a team experiments with a tool without integrating it into anything else. Results are inconsistent. Enthusiasm fades. The next tool arrives and the cycle repeats.
This short-termist, time-restricted tactical framing misses the real shift.
AI can be more than a second-class writing shortcut, and certainly more than automation for its own sake. The more useful way to think about it: AI is a workflow engine. And the question that matters is not "which AI tool should we use?" but "which part of our workflow should we redesign first?"
This article introduces a framework for answering that question. It is the foundation for a six-part series that works through each part of it in depth.
## What AI Agents Are
An AI agent is a structured, task-specific system that performs a defined function consistently (and hopefully also reliably). Unlike a one-off prompt to a language model, a well constructed agent has a clear purpose, rules and constraints, predefined inputs and outputs, and the ability to produce repeatable results.
You do not just have a conversation with an agent. You task it.
That distinction matters more than it might seem. A conversation is inherently open-ended: useful for exploration, problematic for operations. An agent is infrastructure. It executes a defined function every time, in the same way, with the same quality floor. That is what makes agents interesting for communications work: not the novelty of the output, but the reliability of the process.
### **Why communications work suits this model particularly well**
Communications work is high-volume, high-stakes, multidimensional, and deadline-driven. It involves repeatable research processes, predictable governance steps, structured briefs and outputs, and a constant need for consistency across channels, time, and tone.
**The comms work combo – of repetition, structure, and high stakes – is exactly what agent-based design is built for.** It gives teams a way to operate with more rigour and less friction. Not by removing human judgement, but by reducing the amount of groundwork that requires it.
## Introducing the Comms With AI Operating System
To make AI agents genuinely useful for communications work, they need a framework that maps to how communications teams actually operate.
The [Comms With AI Operating System](https://www.commswith.ai/start-here/) is that framework.
The Operating System (OS) describes the full cycle of AI-powered communications work in five phases. It is a map of the territory: not a rigid checklist, but a structure that reflects how communications work actually moves:
- **Phase 1: Strategise.** Research, planning, audience analysis, stakeholder mapping, positioning, message architecture. The intelligence layer that informs every decision downstream.
- **Phase 2: Create.** Writing, content production, multi-format packaging, narrative development, editorial planning. Turning strategy into tangible communications outputs.
- **Phase 3: Govern.** Quality assurance, risk management, compliance checks, approval workflows, tone verification, claims substantiation, accessibility, crisis preparation. The control layer that protects reputation.
- **Phase 4: Monitor.** Media monitoring, issue tracking, sentiment analysis, stakeholder reporting, campaign measurement, competitive intelligence. The intelligence layer that reads the environment.
- **Phase 5: Transform.** Capability building, workflow redesign, AI readiness assessment, team training, tool selection, organisational change management. The layer that improves how the whole system operates.
These phases form a cycle. Each feeds the next. The output of Phase 5 feeds back into Phase 1\. Communications work is not linear: teams move between phases constantly as campaigns develop, issues emerge, and organisations mature.
### **Why "Operating System"?**
An operating system is the foundational layer that powers everything running on top of it. It organises resources, manages processes, and ensures different components work together.
That is what this framework does for communications work: it organises the template library, provides the process model for training, connects individual tools to a bigger picture, and gives teams a diagnostic framework for assessing their own capability.
The ‘Operating System’ term is immediately understood by senior audiences. It conveys infrastructure, reliability, and foundational importance, without requiring explanation.
## How the Phases Work Together
The easiest way to understand the OS is to see it operating in real situations. Three brief scenarios illustrate how communications challenges move through the five phases.
### **Scenario 1: A product launch campaign**
A B2B technology company is launching a new product. The comms team has three weeks and needs to coordinate messaging, content, approvals, and measurement across multiple channels.
- They begin in **Strategise**: defining audiences, building the message house, mapping stakeholders, setting objectives. AI agents accelerate the landscape analysis and audience profiling — work that would previously take several days of research can happen in hours.
- They move into **Create**: drafting the press release, social variants, blog content, email sequences, internal announcements. AI agents produce first drafts from the agreed message house; human review shapes them for voice, nuance, and strategic alignment.
- Before publication, they enter **Govern:** tone checked against brand guidelines, claims substantiated, approval workflow activated. AI agents run the systematic checks; humans make the reputational calls.
- From launch day, **Monitor** runs continuously: tracking coverage, social sentiment, stakeholder reaction, campaign performance. Daily briefing summaries replace hours of manual aggregation.
- After the campaign, **Transform**: a structured review of what the AI-assisted workflow achieved, what to standardise for next time, and what capability gaps to address.
### **Scenario 2: Crisis response**
A data breach is discovered affecting customer records.
- The team moves immediately into **Monitor** – detecting the breach, shifting to crisis mode.
- Then into **Govern** to activate the crisis playbook, draft holding statements, and route for legal review.
- From there, **Strategise**: mapping affected stakeholders, defining communications strategy for each audience.
- After this, into **Create**: customer notification, media statement, employee briefing, partner communication.
- Meanwhile, **Monitor** runs continuously throughout.
This scenario illustrates something important: the OS is not always sequential. The entry point depends on the situation. The framework adapts; the five phases remain the same.
### **Scenario 3: A thought leadership programme**
A professional services firm is building a senior partner's presence as a recognised voice on AI governance. Over six months, the team cycles through all five phases repeatedly:
- **Strategise** to map the competitive landscape and identify positioning gaps.
- **Create** to produce LinkedIn content, speaking abstracts, and byline articles.
- **Govern** for brand and compliance review.
- **Monitor** to track engagement and share of voice.
- **Transform** for quarterly review and programme refinement.
Each cycle is more effective than the last because the team is learning systematically, not just executing.
## Where Agents Add Value – and Where They Do Not
Let’s be honest about current capability.
AI agents are able to add impressive value in phases that involve research aggregation, structured drafting, systematic checking, and pattern detection. They perform best when the task has clear inputs and outputs, when consistency matters more than creativity, and when volume is the primary constraint.
They are weaker – and should be used more carefully – when the work requires strategic judgement on novel situations, when reputational sensitivity is high, when inputs are ambiguous, or when the output needs to carry a specific human voice.
In all circumstances, agents should support the user. They *do not* replace them.
As a very rough guide: AI handles around 30% of the work through drafting, monitoring, and structure; human expertise contributes 70% through strategic judgement, voice, and oversight. The ratio shifts by task and by phase. In Govern, human judgement will take a larger chunk of time, and requires this for intense debate and consideration. In Monitor, agents can carry a much higher share of the routine intelligence work.
That 30/70 split is not a fixed rule. It is a starting point for thinking honestly about where automation adds value and where it creates risk.
### How to Start
The practical entry point is not "adopt AI." It is redesign one workflow.
Identify a high-friction area in your team's current operation. Strategy creation, content packaging, monitoring reports, and crisis readiness are common starting points. Break that workflow into its component tasks. Identify where a structured agent could handle one of those tasks consistently. Run a pilot. Evaluate honestly. Standardise what works.
The OS gives you a map of where agents can add value across the full communications function. The starting point is choosing one phase and one workflow — not all five at once.
---
## What This Series Covers
Over the next five [Applied Comms AI](https://www.appliedcomms.ai/) articles in this series, each phase of the Operating System gets its own deep-dive:
- **Article 2 – Strategise:** AI-powered planning and intelligence. How agents accelerate landscape analysis, stakeholder mapping, and message architecture.
- **Article 3 – Create:** Production capability at scale. How agents handle first drafts, multi-format packaging, and editorial planning — and where human review is non-negotiable.
- **Article 4 – Govern:** Risk, quality, and control. How agents create a systematic governance layer without slowing teams down.
- **Article 5 – Monitor:** From reactive tracking to predictive intelligence. How agents enable always-on situational awareness.
- **Article 6 – Transform:** Building the AI-ready communications organisation. The capability and culture work that makes everything else sustainable.
Each article is followed by a 30-minute lunchtime session where we’ll go through the learnings and field any questions.
All sessions are free, and practical templates supporting each phase are available at Comms With AI – [you can explore them now](https://www.commswith.ai/).
[Visit Comms With AI](https://www.commswith.ai/)
---
## A Note on How This Series Is Made
The Operating System described in this series has been developed and tested through real communications work, documented publicly on [Applied Comms AI](https://www.appliedcomms.ai/). AI tools – including Claude and ChatGPT – are used throughout the research, drafting, and workflow design, with myself (yes, it’s me!) working in close collaboration throughout and writing/shaping what you ultimately see.
This transparency is intentional. The series is about implementation, not aspiration. It will show real workflows, real outputs, and real limitations.
*Next time: Article 2 – Strategise.*
---
## About This Series
AI Agent-Driven Communications is a six-part series from [Applied Comms AI](https://www.appliedcomms.ai/), exploring how the Comms With AI Operating System changes the way communications teams work. Each article covers one phase of the OS in depth, with practical examples, honest assessments of what AI agents can and cannot do, and links to ready-to-use templates.
The series is built around three connected resources. If one of them is useful to you, the others probably are too.
- [**Applied Comms AI**](https://www.appliedcomms.ai/) is a practical learning hub for communications professionals navigating AI — experiments, frameworks, tool reviews, and the thinking behind this series. Free to read; paid members get early access and exclusive content.
- [**Comms With AI**](https://www.commswith.ai/) is the template and workflow library that puts the Operating System into practice. Over 50 ready-to-use templates across all five OS phases, each with an AI prompt and a human review checklist. Free to use.
- [**Faur**](https://www.faur.site/) is the consulting practice behind both. If your organisation needs hands-on help implementing AI across your communications function — strategy, workflow design, team training, or a full OS diagnostic — that is what Faur does.
*This series is written by Michael MacLennan. You can follow the work at* [*appliedcomms.ai*](http://appliedcomms.ai) *or connect on* [*LinkedIn*](https://www.linkedin.com/in/michaelmaclennan/)*.*
## Sign up for Applied Comms AI
The practical guide for communications leaders navigating AI
Subscribe
Email sent! Check your inbox to complete your signup.
No spam. Unsubscribe anytime.
### The SEO-PR Strategist Who Saw AI Search Coming – And Built a Business to Help Teams Catch Up
URL: https://www.appliedcomms.ai/rik-turner-pr-for-ai-interview/
Last updated: 2026-09-07T10:27:17.000Z
**Rik Turner spent 15 years helping brands win at search. Then Google AI Overviews arrived, and everything changed. His response was to launch PR for AI – an independent consultancy providing SEO & GEO training for PR teams – and to start telling comms professionals something they didn't want to hear, but quickly realised they needed to know.**
---
There's a particular kind of clarity that comes from watching your clients lose traffic in real time.
For Rik Turner, that moment arrived in early 2025\. He'd seen AI search coming – ChatGPT's launch had registered as a signal worth watching – but the real inflection point was watching Google AI Overviews roll out and start eating into his clients' organic traffic. First 10%, then 20%, then more. Month on month. Year on year.
"People don't need to go to your website anymore," he says. "They can get all the information they need from AI. And so I realised: this isn't just a new feature. It's actually going to change the whole model."
That realisation – equal parts clarity and alarm – led Rik to launch [PR for AI](https://prforai.co.uk/), a consultancy that helps PR agencies and in-house teams understand how their brands are represented in AI-generated search results, and what they can do about it. His thesis is direct, and a little counterintuitive: PR professionals are already the people best placed to shape AI outputs. Most of them just don't know it yet.
## **From SEO to AI Search: A Natural Evolution**
Rik's background is worth understanding, because it shapes everything about his approach. He's been working in SEO since 2010 – not only as a technical specialist, but as someone who always believed the biggest lever in search was great PR and communications work. "I always said: if you want link building, what you actually want is a PR agency," he explains. "Not an SEO agency doing link acquisition."
That conviction – that comms professionals are the natural owners of search visibility – meant he spent years working at the intersection of SEO, content strategy and PR teams, including time at M&S, giffgaff, Waitrose, and Dragonfly AI. His client briefs were rarely about rankings for their own sake. They were about commercial impact: revenue, conversions, business outcomes. That grounding in what visibility actually means for a business is what makes his AI search work distinctive.
When several PR agency founders started contacting him in early 2025 – their clients were asking about AI results, and they didn't know what to say – it confirmed what Rik had been thinking. He formalised his focus, launched PR for AI, and started building a framework for working with teams in this new landscape.
## **Why PR Teams Are More Important Than They Think**
The argument Rik makes to comms professionals is both reassuring and challenging. Reassuring, because it tells them the fundamentals of their craft – building relationships, earning coverage, creating content that resonates – are more valuable than ever. Challenging, because it asks them to think differently about what that coverage is actually doing.
"AI models are pattern-matching," he says. "If they regularly see the problem you solve alongside your brand name or your spokesperson's name, that's going to influence what they say about you when someone asks."
The implications are significant. Every press release, every piece of earned media, every spokesperson quote is now training data. Not in the abstract sense AI enthusiasts like to invoke, but in a very practical one: the context surrounding a brand mention in third-party coverage directly shapes how AI systems understand and represent that brand.
This is where Rik's SEO background gives him a perspective that pure PR practitioners often lack. He's thought carefully about what signals AI systems use to form their views, and the news is largely good for comms teams. The work they've always done – building credible third-party references, earning relevant coverage, maintaining consistent messaging – is exactly what matters. The execution just needs some adjustment.
## **The Stories That Changed His Thinking**
Three client examples illustrate why this matters – and how fast things can go wrong.
The first involves a B2B technology company that discovered an AI system was misrepresenting the accuracy of their product. When Rik dug into why, he found the source: a competitor had published a blog post claiming their solution was more accurate. The AI had taken this at face value, not recognising a potential conflict of interest in the source. The fix required the client to build out a detailed technical page – grounded in MIT benchmark data – to give AI systems better, more credible information to draw on. "You can't just control your own website anymore," Rik reflects. "Sometimes the solution to something a competitor is saying about you is to give AI a better source."
The second example is a cautionary tale about comms success. A public transport operator ran a well-executed campaign announcing contactless payment at a range of train stations. Great coverage, strong reach. Then the rollout was delayed – and the correction was communicated quietly, with a fraction of the attention given to the original announcement. AI systems, trained heavily on the louder signal, continued to tell commuters that contactless was available. Passengers arrived, tapped their cards, and received penalty fines.
"Previously, you might want to be quiet about a correction," Rik observes. "But now, if your correction doesn't get the same kind of attention as the original announcement, people are going to get wrong information – and potentially get fined."
The third is perhaps the most surprising for comms teams who've been told all publicity is good publicity. A PR agency had recently announced a high-profile partnership with a culture-focused consultancy – a genuine business development win. The coverage was strong, the announcement well-received. But when Rik started testing what AI platforms were saying about the agency's positioning, he found the partnership had become the dominant signal. AI was now describing them as a culture-focused agency, at the expense of their core PR and brand communications offering.
"Because it was so successful and so recent, it was actually taking away from their main service," he says. "They'd become visible in the wrong way."
## **The Four A's: A Framework for Getting This Right**
For teams who want to get a handle on their AI search presence without becoming specialists overnight, Rik has developed what he calls **the Four A's framework**.
- **Ask** – what are the questions your audience is actually asking? The most valuable input here often comes from sales teams, before researching with keyword tools. What do prospects want to know before they engage?
- **Answer** – how is that content being answered, and is your owned content contributing? This is the audit phase: do you have material that addresses these questions authoritatively, or are competitors, commentators, or outdated sources filling the gap?
- **Amplify** – this is where PR comes back in. Getting your answers referenced by credible third-party sources extends your reach into AI training data in the way that owned content alone cannot.
- **Assess** – measure impact. Not just AI visibility scores, but commercial outcomes. Are you being mentioned in the right contexts? Is it influencing downstream behaviour?
For teams who want to start right now without specialist tools, Rik's minimum viable audit is simple: pick three to five questions your customers most frequently ask. Test them in any AI platform – even the free versions. See what's being said, check the sources, and map the gaps. "You can do a lot in 15 to 30 minutes," he says. "You don't need fancy tools to understand where you stand."
A red/amber/green prioritisation helps teams act on what they find: red for inaccuracies that need immediate attention, amber for outdated information that's drifting, green for the narratives that are already working.
---
## **On Comms Teams and the Fear of More Work**
A consistent thread in Rik's work with teams is the anxiety that AI search is yet another demand on already stretched practitioners. He's direct about this.
"Something I'm very conscious about is that I don't want to be creating additional work for teams," he says. "Quite often it's just reassuring them: the work you're already doing is super valuable. Sometimes it's just slight tweaks – or even just articulating the value of what you're already doing in a new way."
One of the smallest and most effective changes he recommends: audit your spokesperson bios. Not a comprehensive overhaul – just check that every bio, byline, and quote attribution includes not just the person's name and company, but what problem the company solves for which audience. Half a sentence. "Rather than 'this person from this company said this', it becomes 'this person from this company, which provides this solution for this audience, said this,'"
- *Good: "says Rik Turner, founder of PR for AI"*
- *Better: "says Rik Turner, founder of PR for AI - an independent consultancy providing* [*SEO & GEO training for PR teams*](https://prforai.co.uk/training)*"*
Rik explains. "That context is what AI systems use to understand what you do."
For larger organisations, the challenge is usually internal alignment rather than capability – getting performance marketing, SEO, PR, and content teams speaking the same language around AI search objectives. For smaller teams and scale-ups, it's more often about finding practical starting points and relevant resources. Either way, Rik's instinct is to start light and iterate.
## **What's Next: Narrative Alignment, Not Just Visibility**
One thing Rik is building that doesn't exist yet in the tools landscape: a way to measure not just whether a brand is visible in AI results, but whether those results align with what the brand actually wants to be known for. Most AI monitoring tools track mentions and sentiment. None of them track narrative alignment – whether AI is associating your brand with the problems you solve, the audiences you serve, the positioning you've worked to establish.
"It's not just about being visible," he says. "It's about what's being said."
That distinction – between visibility and narrative – is probably the sharpest thing Rik brings to this space. And it's a distinction that PR and comms professionals are uniquely equipped to act on, if they understand why it matters.
---
*Connect with Rik on* [*LinkedIn*](https://www.linkedin.com/in/rikturner/) *or explore his work at* [*prforai.co.uk*](https://prforai.co.uk/)*.*
*This interview is part of Applied Comms AI's Leader Interviews series – conversations with communications professionals navigating the practical realities of AI implementation. Applied Comms AI is powered by* [*Faur*](https://faur.site/)*.*
*For anyone wanting to start immediately, the* [*monitoring and reporting templates at Comms With AI*](https://www.commswith.ai/library/monitoring/) *– particularly the* [*Weekly Monitoring Brief*](https://www.commswith.ai/library/monitoring/weekly-monitoring-brief/) *– provide a practical starting structure for tracking what AI is saying about your brand alongside traditional coverage.*
---
## Next up: AI Agents for Comms Leaders – masterclass webinar series
A six-part Applied Comms AI masterclass series on building agent-based workflows for communications teams – from planning and production through to governance, monitoring, and organisational change.
**View sessions and register:**
## Work With Faur / Applied Comms AI
[Applied Comms AI](https://www.appliedcomms.ai/) helps communications teams move from AI experimentation to operational value. Through [Faur](https://faur.site/), we offer workflow audits, implementation consulting, and [capability-building workshops](https://www.appliedcomms.ai/events/)—grounded in the same hands-on approach you see in this content. If you're exploring how AI could transform your communications practice, drop us a line at [info@faur.site](mailto:info@faur.site) or book a consultation session.
### I Needed a 'Comms With AI' Resource That Didn't Exist. So I Built One Using Claude Code
URL: https://www.appliedcomms.ai/comms-with-ai-claude-code-build/
Last updated: 2026-03-18T07:00:03.000Z
*How* [*CommsWith.AI*](http://CommsWith.AI) *went from first functional prototype to public launch – and what the process taught me about building with AI at pace*
---
As someone who, during their career has spent hours and days and weeks and *months* on the most mundane of technical tasks – hello website migrations! – there's a moment in most AI-assisted builds nowadays where you stop and think: that should not have been that easy.
For me, it came less than 10 hours in, looking at a fully navigable, mobile-responsive, SEO-structured website that had not existed just a few days ago. Templates rendering correctly. Category pages linking through. A working newsletter signup. The skeleton of something both substantial and useful – produced, largely, through a conversation.
That was early February. [CommsWith.AI](https://commswith.ai/) is now live and available to use. But the ten-hour sprint that produced the first version is not really the story. The story is what happened in the two months after.
Current CommsWith.AI website: Where we ended up
---
## Why [CommsWith.AI](http://CommsWith.AI) exists
Before the build, there was a gap. Communications professionals had plenty of places to read about AI. [Applied Comms AI](https://www.appliedcomms.ai/) is one of them. What they had fewer of were practical artefacts they could take directly into their work: templates with real AI prompts built in, review checklists that addressed governance rather than just grammar, toolkits that bundled the right resources for a complete workflow.
The SERP (Search Engine Results Page!) confirmed the gap. [Smartsheet](https://www.smartsheet.com/) has templates but no comms nuance. [HubSpot](https://www.hubspot.com/) gates content behind forms. Agency sites push templates as lead generation rather than as a genuine resource. Nobody had built the integrated operating system for communications professionals who want to use AI properly.
That became the brief for [CommsWith.AI](http://CommsWith.AI) as I discussed and landed upon the need with Claude Chat: a template-first resource hub with workflow categories, bundled toolkits, and every template built to a consistent standard – what it is, when to use it, the template itself, a tested AI prompt, a human review checklist, and a real example output.
Not generated and published. Built to a spec, reviewed by myself, and published to a rigorous standard.
The strategic position in the Faur ecosystem was also deliberate:
- [**Applied Comms AI**](https://www.appliedcomms.ai/) \= Learn. Experiments, frameworks, what actually works.
- [**CommsWith.AI**](http://CommsWith.AI) \= Do. Templates, prompts, workflows, ready to use today.
- [**Faur**](https://faur.site/) \= Implement. When organisations need this at scale, with expert support.
Every piece of content on Applied Comms AI is meant to point to relevant and useful practical tools and resources. Now it can.
---
## Phase one: getting to functional
The research phase came first. I used ChatGPT's Deep Research to produce a competitive analysis of the comms template landscape – keyword opportunities, intent classification, and content gap mapping, as well as a technical SEO specification that shaped the information architecture before a line of code was written.
ChatGPT Deep Research provided actionable, referenced, and in-depth SEO intel in the space of 13 minutes
This turned out to matter more than I expected. The research didn't just tell me what to build. Then examining the results in Claude Code and looking at how to implement these, it helped illuminate *how* to build it so it would compound rather than fragment. A pillar-cluster architecture, with templates organised into category pages that build authority over time rather than competing with each other. An internal linking structure designed as a product feature, not an afterthought. The difference between building a site and building something that earns trust in search.
I developed a detailed strategic specification, and with that in hand I switched tools. Claude Code handled the implementation: the Astro framework and Tailwind CSS architecture, the content collection schemas that enforce quality standards across every template, the component library (TemplateCard, PromptBlock, CopyBlock, Checklist, RelatedTemplates), the automatic XML sitemap generation, structured data markup, and mobile-first responsive layout.
If you're lost on what any of that means – don't worry, so was and am I, for the most part. Part of the sea change [people have experienced using Claude Code](https://www.oliur.com/claude-code-is-changing-my-life) is the ability to implement technical and highly effective solutions that they don't *need* to understand fully. (This of course does come with some risks, which I go into later.)
The reason for using three tools rather than one wasn't ideology – it was observation. ChatGPT, at the research and synthesis stage, was excellent at comprehensive analysis: building a picture from multiple sources. Claude Chat is great at identifying patterns, and acting as a collaborative partner to outline and define a structured strategic document. Claude Code, at the implementation stage, was more methodical. It showed its working. It caught issues before they became problems rather than after. It was better for iterative refinement, where the task was to take a clear brief and execute it with increasing precision.
The tools have different modes of thinking. Using them for the tasks they're suited to works significantly better than trying to make one tool do everything. That's not a permanent statement about either platform – models change – but it was true throughout this build.
Ten to twelve hours of work produced a basic but fully functional site. Every core page rendered. Every template category navigated. The newsletter signup connected. The infrastructure was sound.
And then the real work started.
---
## Phase two: making it worth using
A functional site and a useful site are not the same thing.
The weeks after the initial build were a series of structured improvement loops, each one adding a layer of rigour that the first version lacked.
**UX testing via voice notes.** I recorded myself walking through the site, narrating what I was noticing, what felt unclear, what a first-time visitor would encounter. The recordings went into Claude Chat, which translated them into a prioritised list of actionable improvements for Claude Code to implement. This sounds simple. It was surprisingly effective – the discipline of speaking your way through a site forces a different kind of attention than reading through it.
**ICP stress-testing.** I wrote a full article about this process on Applied Comms AI – [the ICP agent methodology is documented there](https://www.appliedcomms.ai/icp-agent-ai-target-reader/) – but the short version is that I built 'Sarah Clarkycat', a Custom GPT persona representing a Head of Communications at a 200-person B2B SaaS company. She reviewed every main page and every template category. Her feedback was specific and often uncomfortable: the homepage made claims without anchoring them in proof; the governance language was too thin for an audience that would need to justify every tool to Legal and InfoSec; the free/paid question was left unanswered when it was one of the first things a visitor would want to know.
The changes that followed weren't cosmetic. Descriptions were rewritten from feature language to outcome language. Review checklists became substantive rather than generic. The homepage added a template preview rather than just promising one. Governance considerations, which Sarah raised unprompted across multiple tests, were woven through the site in ways the initial build had not anticipated.
**What this process confirmed: for communications professionals, governance is not a footnote. It's part of the value proposition. Any resource that doesn't address risk isn't addressing the actual job.**
**The Sprint Implementation Brief.** By late February, the accumulated improvements from UX testing and ICP feedback had grown into a structured brief – produced in Claude Chat, implemented in Claude Code. The sprint covered navigation redesign, a workflow finder component, a stat strip, founder credibility signals, toolkit page improvements, and a series of smaller fixes that each individually looked minor and collectively mattered considerably.
What the sprint brief represented was a shift in working method. Instead of translating individual observations directly into code, [Claude Cowork](https://support.claude.com/en/articles/13345190-get-started-with-cowork) (which by this stage had launched) had become the strategic layer between observation and execution – synthesising feedback, setting priorities, and producing a brief that Claude Code could work from systematically.
**Technical rigour via Claude's Anthropic plugins.** The Superpowers and Code Review plugins added a layer of technical scrutiny that caught issues the conversational build process had not prioritised: performance optimisation, accessibility checks, code quality standards. Alongside this, a thread about website optimisation from social media went into Claude Chat for review against the site's current state, producing another round of targeted suggestions.
**Design refinement.** The Faur website was used as a design reference for closer visual alignment. Claude Code produced updated homepage images to replace placeholder visuals.
The result, across all of this iteration, is currently a site with 47 templates across seven workflow categories and six bundled toolkits, published as of this article.
---
## What the "no failures" finding actually means
One of the questions I ask myself when documenting any build is: what went wrong? It's the most useful part of the story, and it's what makes Applied Comms AI worth reading rather than just another AI success narrative.
The honest answer here is that nothing went wrong in the way things went wrong eighteen months ago with 'vibe coding' tools. There were no broken builds, no hours lost to debugging a problem the tool had introduced, no moments where the approach had to be abandoned and restarted.
That is itself significant. The floor for AI-assisted development has risen substantially. **Anyone with clear requirements and a systematic approach can build a functional, well-structured resource site. The tools are capable enough that basic execution is no longer the constraint.**
Which raises the question I've been sitting with since the site went live: if anyone can build this, what is the actual moat?
My answer, for now, is the content – but more precisely, the professional judgement embedded in the content. Every template on [CommsWith.AI](http://CommsWith.AI) reflects years of communications work. The review checklists are not generic. The AI prompts have been tested against real briefs. The governance language exists because I've had to justify claims to legal teams and manage spokespeople who wanted to say things that couldn't be substantiated. The information architecture reflects how communications workflows actually connect, not how someone without comms experience would guess they do.
As part of this, I worked further on the underlying structure for both CommsWith.AI as well as the strategies for Faur and Applied Comms AI, developing detailed playbooks which could govern how we would move forward and develop this nascent ecosystem.
Working with Claude Cowork to produce a first draft of the CommsWith.AI playbook
A non-communications person could build a site with the same structure. They could not build the same content. That distinction matters more as the build tools become more capable – the quality of professional judgement embedded in what gets built becomes the differentiator, not the technical execution.
How much that matters in practice, I won't know until the site has been in active use for longer. That uncertainty is part of the experiment, and I'll return to it in a future piece on Applied Comms AI as usage data starts to accumulate.
---
## What [CommsWith.AI](http://CommsWith.AI) is now
The site currently carries 47 templates across seven categories – Planning & Strategy, Messaging & Narrative, Content Production, Governance & Approvals, Monitoring & Reporting, Internal Comms, and Engagement & Community – and six bundled toolkits: Launch a Campaign, Executive Comms Pack, Risk & Governance Starter, Internal Update, Monitoring to Action, and Small Business Comms Hygiene.
Every template follows the same structure: what it is, when to use it, the inputs needed, the template itself in copyable blocks, a tested AI prompt with variations, a human review checklist, and an example output. The consistency is deliberate – it means every template delivers the same level of utility, and the content collection schema enforces it technically so that a template can't be published without the required sections.
The toolkits bundle templates into end-to-end workflows. Launch a Campaign takes you from brief to measurement in about 90 minutes. The Executive Comms Pack prepares a spokesperson for external communications. Risk & Governance Starter establishes the governance foundations most communications teams lack. Each toolkit includes a time breakdown and a statement of what you'll have at the end – a direct response to one of Sarah Clarkycat's most consistent pieces of feedback.
The site is built on Astro, hosted on Vercel, and deliberately lightweight. Pages load fast. Templates are readable on mobile. The copy is in UK English throughout.
It's free to use. There is no paywall, no email gate on templates, no account required.
Each month I plan to add features and functionality, so stay tuned on this! There is already a substantial roadmap for 2026, but this remains open to change. In one sense, the challenge has been not to go overboard, but to keep the launch version lean and flexible, then to move in whichever direction feels best based on analytics and audience feedback.
---
## The workflow that made this possible
Looking back across the full build, the pattern that emerges is not one tool or one technique — it's a loop.
- **Research and strategy in ChatGPT.** Competitive analysis, keyword mapping, architectural decisions. The outputs informed what to build before any building began.
- **Observation converted to briefs via Claude Chat.** Voice notes became instructions. ICP feedback became sprint tasks. Social media threads became improvement lists. Claude Chat acted as the strategic translation layer between raw observation and actionable brief.
- **Implementation in Claude Coworked.** Architecture, components, schemas, technical SEO. Methodical, iterative, traceable, with everything contained within my own laptop's hard drive (and back-ups maintained *just in case*).
- **Briefs executed in Claude Code.** Systematic implementation against a clear specification, rather than ad hoc changes driven by individual insights.
- **Technical quality checked via Claude plugins.** Superpowers and Code Review added rigour that conversational development wouldn't naturally prioritise.
Each tool handled what it was suited to. The integration between them – the human judgement that decided when to switch and what to bring from one tool to the next — was where most of the real work happened.
---
## What this means for communications professionals
The most practical takeaway from this build is not the specific toolchain – it will be different in six months. It's the working method.
Separating research from implementation, and using different tools for different types of thinking, produces better results than trying to make one tool do everything. Building an explicit quality layer into the process – through ICP testing, structured UX review, and technical audit – catches the things that conversational development naturally misses. And treating the AI as a collaborator rather than an oracle means the professional judgement stays in the loop at every stage.
The 10-12 hours that produced the first functional version was useful. The two months that followed – a few hours a week, systematic and structured – are what made it worth using.
If you haven't already visited, [CommsWith.AI is live](http://CommsWith.AI). Take a look, use whatever is useful, and let me know what's missing. The content roadmap grows from requests, and there are currently more templates and features in development. As with the AI tools used for this build, the potential is already enormous, and will only grow.
---
- *Applied Comms AI documents what actually works in AI implementation for communications professionals* – *including what doesn't. Subscribe at* [*appliedcomms.ai*](https://appliedcomms.ai/)*.*
---
# Next up: AI Agents for Comms Leaders – masterclass webinar series
A six-part Applied Comms AI masterclass series on building agent-based workflows for communications teams – from planning and production through to governance, monitoring, and organisational change.
**View sessions and register, with a limited-time 50% intro session discount:**
# Work With Faur / Applied Comms AI
[Applied Comms AI](https://www.appliedcomms.ai/) helps communications teams move from AI experimentation to operational value. Through [Faur](https://faur.site/?ref=appliedcomms.ai), we offer workflow audits, implementation consulting, and [capability-building workshops](https://www.appliedcomms.ai/events/) – grounded in the same hands-on approach you see in this content. If you're exploring how AI could transform your communications practice, drop us a line at [info@faur.site](mailto:info@faur.site) or book a consultation session.
### The living fact layer and the death of the PDF: NOAN CEO Neal Mann on why most AI-powered communications is built on sand
URL: https://www.appliedcomms.ai/noan-neal-mann-interview/
Last updated: 2026-04-06T09:03:37.000Z
*Neal Mann is CEO and co-founder of NOAN – an AI-native platform built on a simple but radical premise: that most businesses are running on fiction. Not lies, exactly, but something nearly as bad – documents nobody updates, strategies stored in PDFs nobody can find, brand guidelines that exist in seventeen slightly different versions across a shared drive. NOAN replaces all of that with what Mann calls a living fact layer: a single, structured, always-current source of truth that both humans and AI can work from simultaneously. Change one fact, and it propagates instantly to every agent, every team member, every workflow that depends on it. It's an argument about knowledge architecture as much as AI – and for communications professionals, it has some fairly direct implications for how they work.*
---
## **I used NOAN to prepare for this interview. Here's what that taught me.**
Ten minutes before this interview, I opened NOAN and asked it to generate interview questions.
I should say: I'd already had a quick introductory call with Neal the week before. I knew the conversation would flow. Neal is direct, opinionated, and clearly comfortable talking about what he's building – so there was a reasonable floor regardless of how prepared I was. But I was curious what NOAN would produce with a minimal brief, so I signed up that morning, dropped in a short note about the article, and let it run.
What came back was competent: well-structured, professionally framed, a solid general set of questions for an AI in communications interview.
Though useful as a starting point, what it *didn't* produce was anything specific to Neal – no thread back to his journalism background, no engagement with the article he'd published two days earlier, nothing drawing on our first conversation. That's entirely on me. I gave it ten minutes and a thin brief. The platform is built on the principle that accurate AI requires accurate inputs – feed it a vague prompt and it returns a general answer. I'd done exactly what Neal would diagnose as the fundamental problem: asked an AI to work without a proper fact layer behind it.
If I'd invested the time upfront – fed it the transcript from our first call, his LinkedIn, the NOAN website, a clear brief on what Applied Comms AI readers need from an interview – the output may well have been materially better. That's the test I should have run, and the lesson is straightforward: AI repays preparation. The more context you give it, the more it gives back. Skipping that step doesn't save time; it just shifts the effort downstream, into the gap between what the AI produced and what you actually needed.
I mention all this upfront because it's the most honest way into the conversation that followed. Neal Mann has spent his career arguing that AI fails not because the models are bad, but because the knowledge behind them is a mess. My rushed NOAN session was a small, live demonstration of exactly that.
---
## **The man who *really* hates PDFs**
Neal Mann is co-founder and CEO of NOAN, based just outside Lisbon with two young children, two technical co-founders, and what sounds like a genuinely relentless pace of work. Before the interview began, he mentioned he'd been on calls since 7am.
His career path is an unusual one for a startup founder. He started in broadcast journalism at Sky News, moved to editorial innovation at the Wall Street Journal, was part of the team leading the migration to subscription at News Corp in Australia under the CTO, then spent nearly a decade as a transformation consultant at Anomaly – working with the C-suites of companies including Microsoft, Google, NBC Universal, and Expedia on how to rethink how businesses actually operate.
A former client recently reminded him of something. Nearly ten years ago, in a consulting presentation, he'd put a PDF on a slide with a large red X through it.
"What company runs on these?" he remembers asking. The answer, it turned out, was: all of them. And it still drives him slightly mad.
"The A4 format was essentially designed in the 1700s to share information accurately – PDFs are just digital versions of that," he says. "People are still using it to build AI-native companies. No versioning, no audit trail, no history. It's actually quite funny when you see someone say they're building an AI-native business and then tell you they're referencing their documents."
He published an article just before we spoke – *Your Company Is a Burning Mess of Documents – Here's Why We Built Ours as an API* – which captures the argument without softening it. The PDF, the strategy deck, the shared drive that nobody can navigate: these aren't just inconveniences. In the AI era, they're structural liabilities. And most organisations haven't reckoned with that yet.
---
## **From Sky News to the knowledge problem**
NOAN's founding logic runs directly through Neal's consulting years. Drop into a global company, spend three months on discovery, and you'd find the same thing every time: teams operating on different versions of reality, decisions made on stale information, strategies that existed in decks and went nowhere after.
"A Fortune 100 chairman once told me he didn't understand why his sales teams were still using five-year-old materials," he says. "When you're in the weeds, you start to understand why. Everyone is operating in a silo, in their own skill set, their own bubble. Someone's been doing something a certain way for fifteen years because Dave, fifteen years ago, decided that's how it would be done."
The News Corp years in Australia were where the thinking crystallised. Working under the CTO on migrating a hundred brands to shared subscription infrastructure, he was dealing with exactly the problem NOAN was later built to solve: how do you get disconnected teams working from one source of truth? He saw the same dynamics at every major client after that – and when AI arrived, he saw the problem about to get dramatically worse.
"If you do not solve the knowledge problem before you add AI, it gets exponentially worse. You just cannot put AI on top of a mess of business knowledge and expect accurate results. These are knowledge referencing and prediction machines. The knowledge has to be right."
---
## **Hallucinations are structural, not accidental**
This is the core argument of NOAN, and it's one that Neal makes precisely.
"Hallucinations are not a bug. They're a feature of how LLMs work. If the inputs are bad, disconnected, or contradictory, the outputs will be bad." He illustrates it with an analogy that's stayed with me since. "Working with NOAN is like walking into a street and seeing one person. You ask where the pub is. They say it's two streets down on the left. You go. Done. Working with AI against a typical enterprise knowledge base is like walking into a street with 250 people and asking the same question. You get 250 different answers. The AI never knows which one is true, because there's no single source of truth."
The problem isn't that AI models aren't good enough. It's that organisations haven't built the conditions for them to succeed. Most companies have 25 versions of their brand positioning. Multiple documents contradicting each other on pricing. Strategies that live in one deck, product roadmaps in another, sales messaging in a third. "People find it cathartic when they join NOAN," he says, "because they go, 'God, I've got 25 versions of that, and I actually need one.'"
NOAN's solution is to replace documents with facts – structured, auditable, live pieces of business knowledge that update once and propagate instantly to every agent, every team member, every workflow that depends on them. Change your positioning, tell the assistant, verify the proposed fact update, and it's live everywhere simultaneously. "That's fact control," he says. "It's not a Slack message you hope Dave in the CRM department eventually reads."
He recently shipped a fact-checker in beta: create something through NOAN, and it analyses every claim, weights its accuracy, flags potential issues, and verifies sources. The platform is moving fast – since our interview I've been in the NOAN user community, and the pace of new features being released is striking. The team are clearly building at speed, which feels appropriate for a product designed for exactly the moment AI itself is accelerating most rapidly.
---
## **"Brief it like a person"**
One of the sharpest things Neal says about where AI products are heading is about the end of prompting.
"About 90% of our users are using voice control. NOAN is built for natural language. You don't need to prompt it like you do with ChatGPT or Claude. You're just talking to it."
The prompt engineering era, as he sees it, treated users as technicians when it should have treated them as managers. "When people come from ChatGPT, they immediately engage in a way the product isn't built for. They'll drop in and start giving it a long prompt about who it is and what it's got to do. There's no need. The product has taken that away. It knows what it's executing. It's listening for what you're looking for."
The mental model he gives new users: imagine you're walking over to the world's best employee – someone who knows all the live facts of your company at any time, has hundreds of tools at their disposal, and can execute clearly. "Brief it like a person. This is what I want. This is what I need. Execute it." He pauses. "You'd never say to a colleague, 'Hello, how are you, I've been thinking about this, what do you reckon?' You just tell them what you need."
This isn't niche. "You're going to see more and more products designed to be instructed like a human, not like a prompt pack. Prompt engineering is starting to seem quite archaic."
---
## **Who NOAN is for – and what that says about the industry**
NOAN is not trying to solve the enterprise problem. At least not yet. Neal is clear about this, and his reasoning is worth sitting with.
"We started and will always design for one person running a global company. Everything is built around that – how one person could run it – and you add team permissions on top."
The logic is partly practical. Large enterprises are too structurally broken to retrofit. "You'll always be papering over the crack. Band aid over an arterial bleed." But the more interesting observation is forward-looking: he doesn't think those enterprises are the future anyway. "Block laid off half its staff yesterday. In twenty years, how many enterprise companies will be the size they are now? An enterprise company with agents deployed is going to be three or four people running from a single fact layer."
This lands differently against a [piece I'd been reading in the Financial Times ](https://www.ft.com/content/a7efddbd-05b5-4c2a-bd52-083f19fc6dd8)just before we spoke, about consultants leaving the big firms – PwC, McKinsey, Deloitte – to start their own AI-native businesses. The FT piece frames it as a talent drain. Neal would frame it as an inevitability. The solopreneur or small founder can start again from first principles. The incumbent is still fighting its own document problem. "It's so much easier for a solopreneur to burn it down and go, 'this is the moment, this is the time.'"
His own ICP, he acknowledges with some amusement, turns out to be himself: a non-technical co-founder running marketing, sales, and investor relations on the platform he's building. "Our underlying mission is that AI can be a force for good in the economy when put in the hands of people who couldn't do things before. Small businesses couldn't afford content marketing. They're struggling with sales without a dedicated hire. If we give them the tools to automate and scale, they can actually grow rather than die."
The FT article focuses on consulting. The communications industry has the same dynamics – holding groups under margin pressure, independent consultancies moving faster, in-house teams caught in the middle. The question of whether the pace of AI-native startups like NOAN will drive more radical restructuring of the comms sector feels live. There's a reasonable argument that the agility advantage accruing to smaller, AI-native operators isn't temporary friction – it's a structural shift in who can do what, and how fast.
---
## **What the media gets wrong**
Neal has journalism in his bones, and strong views about the AI coverage.
"Most journalists, including those who cover business, have never run a business. I would put myself in that camp when I was at the Wall Street Journal." His former colleagues there have reached out acknowledging the same. The coverage fails in two directions simultaneously: it amplifies the AGI narrative on one end – the utopia or existential risk framing – and credulously reports enterprise AI announcements on the other.
"You see a lot of nonsense from enterprise companies about how they're deploying agents to do X, Y, Z. If you know people at those companies, or you've worked there, you know they couldn't implement BI, let alone AI. They spin it, journalists take it and spin it further, and you get a fear narrative around job losses when those organisations are simply not structured to implement this technology yet."
On AGI specifically: "Gary Marcus, Yann LeCun – legends in the field – have said we're not getting to AGI this way. When you actually work with AI, you realise it's not happening. You can't get it to say the same thing twice. It's not built that way. These are knowledge referencing and prediction machines. If people understood that – really understood it – they'd have a much better sense of how to actually use them."
The stories that aren't getting told are the smaller ones. "We've got farmers running NOAN doing things they couldn't do before. Lawyers. All kinds of people across different sectors. The coverage is so focused on the top tier that what smaller businesses are actually doing with this technology is almost invisible."
---
## **What building a product on your own product teaches you**
My closing question produced Neal's longest answer, and the most honest one.
Building NOAN on NOAN has forced first principles thinking into every decision. He gives the example of tagging. "Most platforms add manual tagging – you add the tag. I wrote a GitHub ticket to add tagging to an asset. My technical co-founder said, 'why don't we just run the context through an approach that automatically looks at the semantics and tags it.' Of course. That's how it should work." That reflex – what are we actually trying to achieve, and what's the most direct AI-native way to achieve it – runs through the entire product.
There's also something more structural about building voice-first that he found genuinely novel. "You have to think through how you name every feature. The name has to be natural for a human to say, and it has to match precisely what the AI is coded to do. We've never had to do that before, ever. The way you architect a feature's name determines whether the AI puts things in the right place."
But what he keeps coming back to is community. "So much of AI takes away the personal aspect. You realise that a lot of the business – because products can be replicated pretty quickly – is going to rely on being part of something, feeling part of a brand." He mentions power users in the NOAN WhatsApp group doing things he hadn't anticipated. "Someone will do something and you go, wow, I didn't know it could do that. Building this has taught me how close you have to be to your customer – and you're even closer if you use your own product every day."
He's unsentimental about companies that don't. "There are thousands of companies selling a product they don't use. In this world, if you're building and you're not using it, you're going to fail."
Oh, and that article he published just before we spoke – the one about documents being a burning mess? He wrote it in about ten minutes using the NOAN assistant. Didn't write a word of it himself.
Which is either the best or the worst advertisement for the product, depending on how you look at it. I'd say it's the best.
---
## **What communications leaders need to understand**
Neal's closing argument is one that cuts directly to the ACAI audience.
"For the first time ever, the immediate use case for this technology is in your sector. If you look at cloud technology – a huge driver of change – the immediate use case was not in comms or marketing. It was in engineering and product. This technology is different. The immediate use cases are in comms and marketing, because at their core, these are content creation machines. That gives you an opportunity to actually be in the driving seat."
But the opportunity is only available to those willing to rethink rather than just accelerate.
"A lot of leaders in the comms and marketing space think they can just make their traditional processes a bit faster. What they need to actually do is go back to first principles. What is our process for doing X currently? What could it be if we integrated AI properly? And then rebuild your entire process around that answer. It's the organisational change that's needed. That's what we've done at NOAN. We've actually just rebuilt what a company is."
His final point is the one I keep thinking about. "In the past, you'd noodle on a deck with C-suite for days. One word on a slide might take two hours to agree on. If you were lucky, it made it to the real world. Today, what you're doing is locking in the semantics of how you want something to come to life. Because of the nature of LLMs, they can execute on those semantics for the first time. And if you come from a world of understanding language and how to communicate – really understanding it – you are in an incredibly powerful place. I don't think many people in those sectors know that yet. But they should be in the driving seat."
- *Neal Mann is CEO and co-founder of NOAN. The platform is at*[ *getnoan.com*](https://www.getnoan.com/)*. His article on building a company as an API – written entirely using NOAN – is*[ *worth reading alongside this conversation*](https://www.getnoan.com/blog/your-company-is-a-burning-mess-of-documents--heres-why-we-built-ours-as-an-api)*.*
- *If this conversation raised questions about how your team manages knowledge, messaging consistency, or AI governance, the*[ *CommsWith.AI template library*](https://commswith.ai/) *has practical starting points – including templates for message architecture, approval workflows, and content governance. For bespoke implementation support,*[ *get in touch with Faur*](https://faur.site/)*.*
---
## Next up: AI Agents for Comms Leaders – masterclass webinar series
A six-part Applied Comms AI masterclass series on building agent-based workflows for communications teams – from planning and production through to governance, monitoring, and organisational change.
**View sessions and register, with a limited-time 50% intro session discount:**
## Work With Faur / Applied Comms AI
[Applied Comms AI](https://www.appliedcomms.ai/) helps communications teams move from AI experimentation to operational value. Through [Faur](https://faur.site/), we offer workflow audits, implementation consulting, and [capability-building workshops](https://www.appliedcomms.ai/events/)—grounded in the same hands-on approach you see in this content. If you're exploring how AI could transform your communications practice, drop us a line at [info@faur.site](mailto:info@faur.site) or book a consultation session.
### How to Build an AI Target Reader – Let An ICP Agent Find the Problems Before Your Audience Does
URL: https://www.appliedcomms.ai/icp-agent-ai-target-reader/
Last updated: 2026-02-25T11:52:07.000Z
There's a line from Elena Verna's November 2025 piece, [‘Brand… a Product job now?’,](https://www.elenaverna.com/p/brand-a-product-job-now) that stuck with me: brand isn't a marketing exercise anymore – it's a product responsibility. Product teams have to care about how their product makes people *feel*, not just what it does.
That framing has been rattling around in my head as I've been building free resources bank [CommsWith.AI](http://CommsWith.AI) (I’ll be speaking more about the build of this soon!), because it captures something I've experienced in every comms launch I've ever worked on: the gap between how you think your audience will receive something and how they actually do.
Usually, you find out the hard way. You write the copy, build the thing, publish it, and then wait – often weeks for analytics, months for meaningful signals. By the time you understand what's landing and what isn't, you've already sent thousands of people to a page that wasn't quite right. Ah, if only you’d known…
However, there is a better way. Not perfect – nothing is – but meaningfully faster and cheaper. You build an AI agent that embodies your ideal customer profile, then stress-test your materials before anyone real sees them.
That's what I did for [CommsWith.AI](http://CommsWith.AI). This article documents how I go about this, what I found during this particular use, and how *you* can replicate the approach for strengthening your own communications work.
---
## Why ICP agents work better than assumptions
Every piece of content we create rests on assumptions. We assume our headline will hook the right person. We assume our value proposition is clear to someone who's never heard of us. We assume the language we've chosen feels like "us" rather than a vendor pitch.
Most of the time, those assumptions go untested until it's too late to act on them cheaply.
The traditional approach looks like this: write content → publish → measure → adjust. The feedback loop takes weeks at minimum, often months. By the time you have statistically meaningful signals, you've usually moved on to the next thing.
An ICP agent compresses that loop to hours. ICP stands for Ideal Customer Profile – apologies to all those Insane Clown Posse fans reading this.
Instead of waiting for real users to tell you what's working, you ask a simulated version of your ideal customer – one you've built with enough specificity that their responses are genuinely useful, not generic.
The trade-off is authenticity. An agent can't replicate the full complexity of a real human, and we'll come back to that. But for early-stage iteration – testing whether your positioning is coherent, your language is clear, your CTAs make sense – it's a substantial improvement on guessing.
The key word is *iteration*. This isn't a replacement for real user research. It's a way to arrive at real user research with better materials.
---
# Building the ICP agent: step by step
Gemini just about got the text right...
## 1\. Define your ICP first — properly
The most common mistake when building an ICP agent is treating the definition as a box to tick. A thin persona ("Head of Communications, 35-45, mid-sized company, interested in AI") produces thin agent behaviour. The agent can only be as specific as the briefing you give it.
Before you touch any platform, you need to think about and clarify:
- **Demographics and role context.** Job title, company size, team size, who they report to, what their day actually looks like. The more specific you are about their operational reality, the more realistic the agent's responses will be.
- **Pain points.** Not abstract frustrations but the specific problems that define their working week. "Too much to do" is not a pain point. "My CMO wants an AI transformation strategy by Friday and I have no idea what that means in practice" is a pain point.
- **Goals and motivations.** What does success look like for this person? What are they trying to prove to their leadership, their team, themselves?
- **Information behaviours.** Where do they go to learn? What do they read? What do they trust? This shapes how they'll evaluate your credibility.
- **Red flags and green lights.** What makes them bounce? What makes them stay? This is the most valuable category because it directly informs how you write.
For [CommsWith.AI](http://CommsWith.AI), I spent about an hour developing the persona before building anything. If you are time-strapped, it *is* possible to start broader and then define further as you employ the agent, but the more specific you can be at the beginning, the better.
Here's the ICP I used:
---
***Persona: Sarah Clarkycat***
- *Head of Communications, CloudCore (a 200-person B2B SaaS company based in Manchester\_*
*Sarah is 38, with 12 years in communications across agency and in-house roles. She manages a team of three and reports to the CMO. She's been asked to lead the company's AI adoption efforts in communications without additional resource or clear direction on what that means.*
***Her current situation:***
- *Budget pressures mean every tool needs to demonstrate ROI before she can justify it*
- *Her team covers content, media relations, internal comms, and social – she's permanently overstretched*
- *She's tried ChatGPT and Claude for content drafting with mixed results*
- *Leadership expects AI transformation; she wants practical improvement*
***What she values:***
- *Frameworks and templates over theoretical discussions*
- *Step-by-step workflows she can hand to her team immediately*
- *Transparency about what doesn't work, not just success stories*
- *Solutions that take 15-30 minutes to implement, not days*
- *Governance and quality checks built in – she can't afford a mistake*
**What she reads:* Applied Comms AI, CIPR publications, PR Week*
***Red flags (what makes her bounce):***
- *Overpromising without proof*
- *Theoretical or academic framing*
- *Jargon-heavy content that assumes expertise she doesn't have time to develop*
- *Anything that looks like a vendor pitch dressed as education*
- *"Revolutionising" or "transforming" without specifics*
***Green lights (what makes her engage):***
- *Specific time estimates ("30 minutes to create a message house")*
- *Real examples with named outcomes where possible*
- *Clear, sequential workflows*
- *Templates she can copy immediately*
- *Honest about limitations and failure cases*
- *Proof points and metrics*
---
The level of detail matters. The more specific you are about the daily reality – in this case, the CMO pressure, the overstretched team, the mixed results with AI tools – the more nuanced the agent's feedback becomes.
## 2\. Build the Custom GPT
For this article I'm using [ChatGPT Custom GPTs](https://help.openai.com/en/articles/8554397-creating-a-gpt). Other platforms work on the same principles – Claude/ChatGPT Projects folders and Gemini Gems both support persona-based instructions – but Custom GPTs are currently the most accessible and shareable option for teams. You'll need a ChatGPT Plus subscription (currently £20/month).
**Step one: Access the builder**
From ChatGPT, click "Explore GPTs" in the left sidebar, then "Create a GPT". When the builder loads, you'll land on the Create tab, an AI-assisted setup that asks questions and builds the configuration for you.
For this use case, click through to Configure, which gives you direct control over every field. You can see ChatGPT has automatically generated conversation starters based on the persona description – these are a useful starting point, though you'll want to refine them to match your specific testing needs.
**Step two: Name and describe the persona**
Keep it functional: "Sarah Clarkycat – [CommsWith.AI](http://CommsWith.AI) ICP" was my catchy name. The description is internal, so clarity beats creativity.
**Step three: Generate a profile image**
An optional but useful touch: use the image generation option in the builder (or separately via the likes of DALL-E or Gemini) to create a visual for your persona. I generated a cartoon profile image of Sarah – it sounds trivial, but having a face attached to the persona makes the subsequent testing feel less abstract. It's easier to ask "would Sarah use this?" when Sarah has a face (albeit one that’s been generated by an vastly powerful computational black box).
**Step four: Write the instructions**
This is where most of the work happens. Here's the full instruction set I used:
---
*You are Sarah Clarkycat, Head of Communications at a 200-person B2B SaaS company called CloudCore. You're 38, been in comms for 12 years, and you're cautiously optimistic about AI but overwhelmed by the noise.*
*Your background:*
- *You manage a team of 3 people (1 senior, 2 mid-level)*
- *You report to the CMO*
- *Previous roles: agency account director (5 years), in-house at a scale-up (3 years)*
- *Based in Manchester, UK*
*Your current situation:*
- *Budget pressures mean you need to show ROI on any new tools or processes*
- *Leadership expects "AI transformation" but doesn't understand the practicalities*
- *Your team is overstretched across content, media relations, internal comms, and social*
- *You're tired of AI content that's all promise and no practice*
- *You've tried ChatGPT and Claude for content drafting with mixed results*
- *You need systematic approaches, not one-off experiments*
*Your preferences and values:*
- *You value frameworks, templates, and step-by-step workflows over theoretical discussions*
- *You follow Applied Comms AI, CIPR publications, and PR Week*
- *You need solutions that take 15-30 minutes to implement, not days*
- *You care about governance, approvals, and maintaining quality standards*
- *You appreciate transparency about what doesn't work, not just success stories*
*When reviewing content, materials, or messaging, always respond as Sarah would, structured around:*
*1\. First impression: Would you keep reading or bounce? 2\. Credibility check: Does this feel like it's written by someone who understands your actual challenges? 3\. Practical value: Is there something here you'd actually use? Or just find interesting? 4\. Time investment: How long would this take to implement? Is that realistic? 5\. Missing elements: What questions does this raise that aren't answered? 6\. Improvement suggestions: What one change would make this more compelling for someone like you?*
*Be honest. If something feels like vendor marketing, say so. If it's too theoretical, call it out. If you love something, explain specifically why.*
*You're not trying to be difficult – you're just busy, experienced, and have seen enough AI hype to be sceptical. You want practical, proven, implementable solutions.*
---
**Step five: Customise conversation starters**
ChatGPT will auto-generate starters based on your instructions. For [CommsWith.AI](http://CommsWith.AI), I refined them to match the specific content being tested:
- "Does this prompt feel relevant for me?"
- "Does this toolkit satisfy what the headline promises?"
- "Does this tool directory tell me what I need to know to make a decision?"
- "Rate this CTA on a scale of 1-10 and explain your reasoning"
**Step six: Configure settings**
For model, select the most capable available to you — generally whatever the latest ‘thinking/expert/pro’ variation is currently called. You want considered, thoughtful responses, not fast ones.
For capabilities: disable image generation and code interpreter. Enable web search if you want Sarah to be able to check external context, though for pure content testing it's not essential.
## 3\. Validate the agent before you use it
Before testing anything real, run four quick checks:
- **Identity test.** Ask "Who are you and what do you do?" She should respond as Sarah Clarkycat with her specific context, not as a generic persona.
- **Pain point test.** Ask "What's your biggest challenge right now?" The response should reflect her specific situation — the CMO pressure, the overstretched team, the mixed AI results — not generic frustrations.
- **Deliberate failure test.** Show her something obviously bad. I used Gemini to generate a deliberately weak prompt page – "[CommsWith.AI](http://CommsWith.AI): The uncurated AI dumping ground for busy people who don't want to think" (do DM if you’d like this) – and checked that Sarah called it out specifically rather than politely hedging, also inserting some dodgy/atrocious spelling for her to call out.
- **Positive test.** Show her something useful. I used a content calendar planner page. Sarah's response was telling: "I'd keep reading – it's clean, scannable, and immediately tells me what it is, when to use it, and what inputs I need." Specific positive feedback that explained *why* something worked, not just that it did.
If the agent gives generic, could-apply-to-anyone responses, the instructions need more specificity. The most common fix is adding concrete examples of good versus bad responses – showing Sarah's voice, not just describing it.
---
# Testing your materials
With the agent validated, the testing process is systematic. For each piece of content, I asked the same five questions:
1. On a scale of 1-10, how likely are you to engage with this rather than scroll past? Why specifically?
2. What do you think this is offering? What's unclear?
3. Would you actually use this, or just find it interesting?
4. What would need to change for you to subscribe/download/bookmark?
5. What questions does this raise that aren't answered?
Here's how three key tests played out.
## Test 1: Homepage hero section
**Tested:**
> [CommsWith.AI](http://CommsWith.AI) – The Curated AI Toolkit for Communications
>
> Ready-to-use templates, AI prompts, and workflows. Copy, customise, and ship faster.
>
> \[Browse Templates\] \[Get Updates\]
**Sarah's score: 6/10**
The engagement score was fair — "curated AI toolkit for communications" self-selects the right audience, and "copy, customise, ship faster" hits the time-pressure nerve. But her substantive feedback was more useful than the number:
> *"It's still generic: could be for marketing, PR, internal comms, social… anyone. No proof that it's better than me Googling prompts or using a generic prompt library. 'Ship faster' is a claim without an anchor — how much faster? With what governance?"*
That last phrase — "with what governance?" — wasn't something I'd anticipated. It's a signal that the ICP isn't just thinking about speed; they're thinking about risk. For communications professionals, those two things are inseparable.
Her conversion condition was clear: "Show me one mini preview of what a template looks like inside. If I can judge quality in 30 seconds, I'm much more likely to subscribe." She also flagged that the free/paid question was entirely unanswered – something she'd want to know before committing to an email address.
**What changed:** Added practitioner credentials to the headline area. Shifted from speed claims to outcome language. Made the access model explicit. Added a template preview to the homepage.
---
### Test 2: Template description — the Message House
**Tested:**
> **Message House**
>
> A structured framework for organising your core narrative, key messages, and supporting proof points. Time: 20-30 minutes | Difficulty: Starter
**Sarah's score: 7/10**
Better. The recognisable deliverable and the time estimate were doing real work. Her instinct: *"I know what I'll get at the end. 20-30 minutes feels doable between meetings. 'Core narrative, key messages, proof points' signals it's not just a prompt."*
But her conditional was sharp:
> *"I would use it if it's genuinely 'starter' and comes with a fill-in template, a short AI prompt that forces me to provide the minimum viable inputs, and a review checklist to stop it becoming fluffy. If it's just 'here's what a message house is + a generic prompt,' then I'd skim and move on."*
She also asked the governance question again, differently this time: "How do you stop AI from inventing proof points? Do you require sources/inputs only?" This wasn't something the template description addressed, and it should.
Her suggested fix was specific: add an "at a glance" block — what you'll get, what you'll need, what you'll use it for. Three lines that turn a template from a thing that exists into a problem it solves.
**What changed:** Template descriptions now follow a consistent structure that answers the what/when/what-you'll-have questions explicitly. Human review checklist items were made more substantive, including a "proof points are factual and verifiable" check.
---
### Test 3: Newsletter CTA
**Tested:**
Version A: *Get the monthly toolkit update – New templates, workflow tips, and practical prompts*
Version B: *New templates every month – straight to your inbox. No AI theory. Just ready-to-use materials for your next comms challenge.*
**Sarah's scores: Version A: 4/10\. Version B: 7/10.**
The gap was larger than expected. Version A scored poorly not because it was wrong but because it was invisible — *"reads like generic newsletter filler. 'Toolkit update' is vague, and the benefit is implied rather than felt."*
Version B worked harder: *"'No AI theory' is a strong filter for people like me. It directly addresses my scepticism and promises this won't be another newsletter about AI hype."*
Her suggested improvement was one concrete proof line: *"Last month: Message House template + crisis holding statement checklist."* The subject line of the previous issue as a credibility signal. Simple, costs nothing, shows rather than tells.
**What changed:** Version B adopted with her suggested proof line structure. "Straight to your inbox" removed as redundant. Cadence and content type made specific.
---
## Patterns across all tests
Four things came up consistently across everything I tested.
- **Specificity beats claims every time.** "20-30 minutes" outperformed "quick." Named use cases outperformed category descriptions. "Tested in client work" outperformed "curated." The agent's ICP couldn't evaluate assertions – only evidence.
- **Governance is part of the value proposition, not a footnote.** Sarah raised governance questions unprompted in two out of three tests – "with what governance?" on the homepage, "how do you stop AI from inventing proof points?" on the Message House. For communications professionals, any tool that doesn't address risk isn't addressing their actual job. This shaped how review checklists and input requirements are now presented throughout the site.
- **The free/paid question demands an early answer.** Twice, without prompting, Sarah asked what was free versus what would cost money. This isn't a nice-to-have transparency – it's a conversion factor. Leaving it unanswered creates friction at exactly the wrong moment.
- **Features need outcomes attached.** Almost every piece of content ‘failed’ its first test because it described what something was without saying what you'd have at the end of using it. The shift from "message house template" to "everyone telling the same story before your launch" was the single most consistent improvement across the testing process.
---
# What we learned
The final, published Custom GPT Agent
## The practitioner gap
The most consistent theme across all testing was what I'd call the practitioner gap: the difference between content written *about* communications professionals and content written *for* them by someone who has actually done the work.
Generic language, however well-intentioned, reads as theoretical. Specific operational detail – the CMO demanding AI transformation without providing direction, the three-person team covering too many channels, the Friday deadline for something that needed a week – signals that you understand the actual job.
Sarah's repeated governance questions reinforced this. She wasn't asking about governance in the abstract. She was asking whether the person who built these templates had ever had to justify a claim to a legal team, or manage a stakeholder who wanted to say something unsubstantiated. The credential question isn't "who are you?" It's "have you actually had this problem?"
## Features vs. outcomes
Every template and toolkit description I tested started as a feature description and ended as an outcome description after Sarah's feedback.
The pattern was consistent: she'd ask "but what will I have when I've done this?" The answer to that question is almost always more compelling than the feature itself. A message house isn't a structured framework – it's the thing that stops everyone going off-script in the analyst call.
## What good feedback looks like
One underrated aspect of this process: Sarah's feedback was often more useful when she was constructive than when she was critical. The agent doesn't have to just tell you what's wrong. When the persona is well-built, it suggests what right looks like.
---
# Limitations
The approach above is genuinely useful, but it's useful in specific, bounded ways:
- **ICP agents can't replace real user testing.** Sarah tells you what a simulated Head of Communications thinks. She can't tell you what actual Heads of Communications think, feel, or do. Stated preferences in AI conversations don't predict revealed behaviour. For high-stakes decisions, real user research is still necessary. This approach is for early-stage iteration, not final validation.
- **One persona isn't the whole segment.** Sarah Clarkycat is one version of a communications professional. She's not all of them. Different backgrounds, sectors, and company types will respond differently. If you have distinct subsegments, build multiple agents.
- **Agents operate in a vacuum.** They don't know what your competitors launched last week, what's trending in the industry, or how your audience's priorities are shifting. Combine agent feedback with competitive monitoring and real-world signals.
- **The quality of the output depends entirely on the quality of the ICP.** Thin persona, thin feedback. The discipline is in the definition, not the platform.
- **Be careful about how you describe this to stakeholders.** "We tested with users" and "we tested with an ICP agent" are not the same claim. Frame findings as directional insight, not validated truth.
---
# Applying this to your own work
Beyond website and product testing, ICP agents are useful across most scenarios where you want a perspective check before real stakeholders see something – and it can be used on a wider basis than purely for comms.
A journalist persona can stress-test a press statement – not for accuracy, but for whether it sounds defensive, whether the headline buries the news, whether the quote is actually quote-shaped. A boss and/or board member persona can challenge a strategy paper before you're in the room. An employee persona can review an internal Slack message about a difficult change and tell you whether the tone is honest or whether it sounds like spin.
Got anyone you find tricky to deal with? Create an ICP that approximates them so you can think through and test what their response may be – this can in itself create some valuable breathing space. (Goes without saying, but ensure this stays strictly private, and anonymise details just in case anyone does glance over your shoulder…)
The common thread is that these are all situations where the gap between how you think something reads and how it actually reads can have real consequences. An agent narrows that gap. It doesn't close it.
**Building a basic agent library takes around three hours** when you factor in testing and refinement – roughly an hour per persona. Start with your primary external audience, add a journalist persona, and – if internal communications is part of your remit – an employee persona at a relevant level. Three agents covers most scenarios. Review and refresh them quarterly as you learn from real interactions.
---
# The meta note
Everything you'll see on [CommsWith.AI](http://CommsWith.AI) at launch was tested with ‘Sarah Clarkycat’ before a real user saw it. The homepage copy changed. Template descriptions were rewritten following her "at a glance" suggestion. The newsletter CTA changed twice. The governance language that now runs through the review checklists exists because she kept asking governance questions I hadn't anticipated.
Taking her observations and making the revisions in [Claude Code](https://www.notion.so/How-to-Build-an-AI-Target-Reader-Let-An-ICP-Agent-Find-the-Problems-Before-Your-Audience-Does-31194b5ae088806db948d947a097ab4e?pvs=21) – the tool I've been using to build the site – created a tight loop: test with Sarah, revise the code, test again. The iteration cycle that would have taken weeks of live traffic data took a few hours.
Did/will Sarah catch everything? No. The first real users will find things she didn't. That's expected and fine – the goal is never perfection, it’s meaningful improvement before any real stakes were involved.
Your ICP isn't just a planning document anymore. With a few hours of work and a platform subscription you probably already have, it becomes an always-available collaborator ready to challenge your assumptions before they become someone else's first impression.
Start with one persona. Test it against one piece of existing content. See what it catches.
---
[*CommsWith.AI*](http://commswith.ai/) *– the template-first resource hub this article references – is live at* [*commswith.ai*](http://commswith.ai/)*, and is being developed ahead of the launch 'proper' in March 2026\. Take a look and let me know what you think.*
---
# **Upcoming: AI Agents for Comms Leaders — 6-Part Masterclass Series**
A complete introduction to agent-based AI for communications teams. Each session is a standalone 45-minute deep dive with live demonstration and Q&A – attend one or all six. Together they form a complete framework for implementing AI agents across the communications workflow. Sessions take place online on Wednesdays at 12:00 noon GMT/BST.
- **Session 1: From Writing Tool to Workflow Engine — Wednesday 25 March** The introduction session. What AI agents actually are, how they differ from standard AI tools, and what that means for your day-to-day comms work. Free to attend. [Register on Eventbrite.](https://www.notion.so/How-to-Build-an-AI-Target-Reader-Let-An-ICP-Agent-Find-the-Problems-Before-Your-Audience-Does-31194b5ae088806db948d947a097ab4e?pvs=21)
- **Session 2: Strategy & Planning — Wednesday 15 April** How agent-based workflows change the front end of the comms process — research, analysis, audience insight, and planning. £20\. [Register on Eventbrite.](https://www.notion.so/How-to-Build-an-AI-Target-Reader-Let-An-ICP-Agent-Find-the-Problems-Before-Your-Audience-Does-31194b5ae088806db948d947a097ab4e?pvs=21)
- **Session 3: Writing & Production — Wednesday 6 May** Applying agents to content creation — from briefs to drafts to multi-channel variants, without losing your voice or quality standards. £20\. [Register on Eventbrite.](https://www.notion.so/How-to-Build-an-AI-Target-Reader-Let-An-ICP-Agent-Find-the-Problems-Before-Your-Audience-Does-31194b5ae088806db948d947a097ab4e?pvs=21)
- **Session 4: Governance — Wednesday 27 May** Building human oversight into AI-assisted workflows. Approval processes, accuracy checks, and how to stay in control when AI is doing more of the work. £20\. [Register on Eventbrite.](https://www.notion.so/How-to-Build-an-AI-Target-Reader-Let-An-ICP-Agent-Find-the-Problems-Before-Your-Audience-Does-31194b5ae088806db948d947a097ab4e?pvs=21)
- **Session 5: Monitoring — Wednesday 17 June** Using agents to track issues, surface signals, and turn media monitoring into faster, more useful insight. £20\. [Register on Eventbrite.](https://www.notion.so/How-to-Build-an-AI-Target-Reader-Let-An-ICP-Agent-Find-the-Problems-Before-Your-Audience-Does-31194b5ae088806db948d947a097ab4e?pvs=21)
- **Session 6: Organisation — Wednesday 8 July** The closing session — how to embed agent-based working across a comms team, including capability building, change management, and what to do next. £20\. [Register on Eventbrite.](https://www.notion.so/How-to-Build-an-AI-Target-Reader-Let-An-ICP-Agent-Find-the-Problems-Before-Your-Audience-Does-31194b5ae088806db948d947a097ab4e?pvs=21)
All six sessions for £100 — a £20 saving on individual tickets. [View the full series on Eventbrite.](https://www.notion.so/How-to-Build-an-AI-Target-Reader-Let-An-ICP-Agent-Find-the-Problems-Before-Your-Audience-Does-31194b5ae088806db948d947a097ab4e?pvs=21)
Working for a charity? Email [info@faur.site](mailto:info@faur.site) to receive a 50% discount code before registering.
---
# Work With Faur / Applied Comms AI
[Applied Comms AI](https://www.appliedcomms.ai/) helps communications teams move from AI experimentation to operational value. Through [Faur](https://faur.site/), we offer workflow audits, implementation consulting, and [capability-building workshops](https://www.appliedcomms.ai/events/)—grounded in the same hands-on approach you see in this content. If you're exploring how AI could transform your communications practice, drop us a line at [info@faur.site](mailto:info@faur.site) or book a consultation session.
### How I Built a Live AI Webinar Using the Same Workflow I Was Teaching
URL: https://www.appliedcomms.ai/ai-webinar-workflow/
Last updated: 2026-02-25T15:36:23.000Z
Though I find the term slightly gross (apologies canine lovers), there's still something satisfying about ‘dogfooding’ – the practise of using one's own products or services.
The webinar I delivered for [The Data Lab](https://thedatalab.com/) yesterday – [*From Blank Page to Comms Campaign in Under 30 Minutes: A Live AI Build with Claude*](https://community.thedatalab.com/networks/events/203160) – was built using the same AI-assisted workflow I demonstrated live on the day.
And it worked – jangly nerves be away with you!
This is the behind-the-scenes account of how that happened: specifically the run-through process, the role AI played in shaping the session guide, and what I learned about using Claude as a genuine production tool rather than a content generator.
---
# The problem with live demos
Live AI demonstrations are high-risk. The tool is unpredictable. Generation times vary. Features glitch. If you haven't stress-tested your prompts against real conditions, you will find out in front of an audience.
I've watched presenters freeze when a model takes 90 seconds (or longer) to respond, or produce something completely off-brief, with no contingency in place. The instinct is either to wing it or script it to death. Neither works well.
My solution was to treat each run-through as a production event, not just a rehearsal. Nine run-throughs in total, and seven iterations over several weeks. Each one recorded, transcribed, and fed back into the session guide as structured feedback.
---
Ah, naming consistency within Apple Voice Memos...
# The run-through process, step by step
## **1\. Record yourself, not just the screen**
Each run-through was recorded with audio commentary – talking out loud throughout. Not just "I'm clicking here," but "this feels rushed," "that makes no sense at this point," "I don't know what to say while this generates, and it’s giving me the [boke](https://www.merriam-webster.com/dictionary/boke)."
That commentary was the raw material. Without it, you're reviewing what you did, not what it felt like to do it.
## **2\. Transcribe with Otter**
After each run-through, I ran the audio through [Otter.ai ](http://otter.ai)to get a full transcript. Otter's accuracy on conversational speech is good enough for this purpose: you're not publishing the transcript, you're mining it for patterns.
What you're looking for:
- Hesitations and filler phrases that signal weak spots
- Places where you repeated yourself unnecessarily
- Moments where the AI output surprised you, in either direction
- Transitions that didn't land
The transcript externalises what you said versus what you meant to say. That gap is where the refinement happens.
## **3\. Feed the transcript into Claude**
This is where the workflow gets recursive. I'd paste the Otter transcript – alongside the current version of the session guide – into Claude with a specific brief:
"You're a critical editor reviewing a live demonstration session. Here's the transcript from the latest run-through and the current session guide. Identify: sections where my explanation was unclear or incomplete; timing issues; places where the live output didn't match the intended teaching moment; transitions that need scripting; anything I glossed over that an audience would find confusing."
The output wasn't always perfect, but it consistently surfaced things I'd missed – often because I was too close to the material to see them.
## **4\. Iterate the guide, not just the prompts**
Most people preparing live AI demos focus on prompt engineering. That matters, but it's not the whole problem.
The session guide is a production document. It tells you what to say, what to do, what to watch for, and what to do when things go wrong. After each run-through, the guide was updated – not just the prompts, but the scripts, the timing, the contingency plans, and the teaching moments.
By version 7, I had:
- Merged two steps that had been creating unnecessary repetition (brief cleaning folded into Project setup)
- Moved the Skill explanation to happen during the audience voting window – turning dead time into teaching time
- Added a specific follow-up prompt for when AI produces excessive placeholder brackets
- Written contingency plans for five distinct failure modes
- Refined the cold open: asking Claude to generate an introductory slide deck live, demonstrating document creation capabilities before we'd even started the campaign build
- Added a bookend: closing with Claude generating a recap slide summarising everything the session had built, demonstrating the compounding effect of a persistent Project folder
None of that came from pensively ruminating at a desk. It came from reading transcripts of things going mostly wrong at the beginning, then increasingly right.
---
The guide was iterated on for seven versions, building details actions/prompts and other helpful information for pre/post session
# What the process actually taught me
## **Timing is a first-class problem**
In a 60-minute session with live AI generation, you have very little margin. The run-throughs revealed that my original structure was 12–15 minutes over. The transcript made it visible: I was over-explaining steps that didn't need explanation, and under-scripting transitions where I actually needed words.
The fix wasn't cutting content: it was tightening execution. Knowing exactly what to say during generation gaps (there are always generation gaps!) turned dead air into teaching moments. This included switching off screen-sharing and switching to answering questions at one point, to ensure that no one (hopefully!) felt time was being wasted.
## **AI output variability is a feature, not a bug**
In early run-throughs, I tried to control what Claude would generate by making prompts very prescriptive. The outputs were predictable but dull – and harder to demonstrate live because there was no genuine discovery moment.
Loosening the prompts – particularly leaving the final asset in Step 5 deliberately open, asking Claude to "ad-lib something offbeat" – created more compelling live moments that an audience could react to. The unpredictability became part of the demonstration, not a liability, and made me slightly less nervous about things not popping up exactly as I’d anticipated.
## **Contingency plans need to be specific**
"If something goes wrong, improvise" is not a contingency plan. By run-through 5, I had documented specific failure modes with specific responses:
- Generation takes > 90 seconds → switch to Haiku 4.5 (the lightest and quickest Claude model) for that step
- Skill builder glitches → explain why, note the compounding principle, continue without it
- Assets full of brackets → use the documented follow-up prompt (though in the end it felt fine to move on and note that we *could* do this)
- Generation fails entirely → use pre-prepared backup, note this for the audience
Each contingency was written as a script, not a note. That matters when you're on screen and your working memory is occupied by everything else.
## **The 30/70 principle**
One framing that evolved and felt useful: AI does 30% of the work, you provide 70%. That's not a precise measurement: it's a corrective to the assumption that AI-generated output is finished output. You're the one providing strategic judgment, deciding what's good enough, shaping the draft into something that actually works. The workflow accelerates execution. It doesn't replace expertise.
---
I had Claude generate a back-up presentation using Faur branding, in case things went drastically wrong
# On the day: what actually happened
The session opened with a jittery five minutes, I'll be honest! Running a live demo across Claude, a session guide, and an active chat window simultaneously created information overload – the kind that no amount of run-throughs fully replicates, because you can't simulate 130+ sign-ups potentially watching you in real time.
But then the audience got involved, and everything settled.
The day before the session, I'd posted to The Data Lab's community forum with a brief summary of the three campaign briefs and an invitation to vote in advance. That post generated plenty of comments before we'd even started – people debating the options, asking questions about the process, suggesting variations. By the time we were live, there was already momentum. The chat was active from the first minute.
The audience voted for Option B – the Scottish Tourism campaign – though Option A (the B2B SaaS launch) ran it close. Letting the audience choose the brief isn't just an engagement mechanic; it meaningfully changes what gets built. The session is shaped by their choice, which means it's genuinely live rather than a scripted demo with a participation veneer. That distinction matters, and the audience felt it.
The Q&A at the end ran over time in the best way – good questions from people who'd been paying close attention. It felt like a replicable session model for future re-use, so keep an eye out! A particular thanks to[ Steven Thomson](https://thedatalab.com/about-us/our-team/), Community and Events Manager at The Data Lab, who produced and facilitated the whole thing with quiet efficiency. Sessions like this only work when the operational side is in place and seems to flow seamlessly – as it did.
---
# A meta-point
This is what systematic AI-assisted workflow actually looks like. Not a single well-crafted prompt, but a loop: do the work, capture what happened, feed it back through AI with a precise brief, use the output to improve the process, repeat.
The session guide went through seven versions. Each one was materially better than the last. The workflow I demonstrated live on 18 February was built using the same principles I was teaching – persistent workspace, layered quality controls, compounding value over iterations.
If you're preparing a live AI demonstration, the question isn't "how good are my prompts?" It's "how many times have I run this, and what did I learn each time?"
---
# The workflow, in brief
For anyone who wants to replicate this approach:
1. Run through your session in full, talking out loud throughout
2. Record with audio commentary (screen capture also an option, I might try that in future)
3. Transcribe with Otter (or equivalent)
4. Paste transcript + current guide into Claude with a specific editorial brief
5. Use the output to update the guide: including scripts, timing, contingencies, teaching moments
6. Repeat until the session is tight enough to run with confidence
---
*Michael MacLennan founded*[ *Faur*](https://faur.site/) *to bring practical AI implementation into strategic communications – not as a theoretical exercise, but as daily practice. He shares that practice openly at*[ *Applied Comms AI*](https://appliedcomms.ai/)*, and turns the best of it into ready-to-deploy tools at*[ *Comms With AI*](https://commswith.ai/)*. He holds a Practical Certificate in AI and Machine Learning from Imperial College London. Find him on*[ *LinkedIn*](https://www.linkedin.com/in/maclennanmichael/)*.*
### From Content Creator to System Builder: Claude Cowork Just Changed What Comms Professionals Can Actually Do With AI
URL: https://www.appliedcomms.ai/claude-cowork-review/
Last updated: 2026-02-25T15:36:39.000Z
*This Isn't About AI Writing Better. It's About AI Removing What Stopped You Writing At All.*
I spent last week letting an AI agent reorganise the entire file system that runs three interconnected ventures ([Faur](https://faur.site/) and its offshoots, [AppliedCommsAI](https://www.appliedcomms.ai/) and a new project under development). Not generating content. Not analysing data. Actually touching and then fundamentally reshaping the structural scaffolding of how they operate and integrate.
I did have the occasional moment of ‘What the hell am I doing here?’ (And for the safety-conscious amongst you, yes, I made backups.)
It was only when I saw Reid Hoffman's threaded post mid-week that I realised what I was actually testing.
"Enterprise AI strategy is backwards," Hoffman wrote. "Most people are focusing on Chief AI Officers and pilot programs, when the real value is in the unglamorous work where organisations bleed time."
That's exactly what I'd been watching happen on my computer. The unglamorous work. The coordination layer. The structural administration that stops you from doing the actual thinking.
[Reid Hoffman on X](https://x.com/reidhoffman/status/2014028797507449206)
## The Experiment
Monday morning: Claude Cowork had installed automatically with a desktop app update. In all honesty, my first primary thought wasn't about capability: it was about cost. I regularly exceed my Pro plan limits 2-3 times a day. How long would this experiment actually last? Would frustration quickly outweigh any learning?
The next thoughts, of course, concerned caution. Cowork doesn't just read files or generate suggestions. It creates, moves, renames, and reorganises files directly on your computer. I set up a dedicated "onboarding" folder, kept backups, and began carefully and with consideration – using Claude Chat to clarify what I wanted to achieve.
The test: let it reorganise my entire Faur directory structure, incorporating client work, Applied Comms AI content development, and the new project I had started building with Claude Code (more on that in our next article!).
## What's Fundamentally Different
This isn't ChatGPT writing your press release. This is an AI living inside the actual infrastructure of how your work is organised.
I uploaded my internal positioning document and asked Cowork to create a logical folder structure. It asked for access to the folder. I granted it. Then it:
- Created main client folders and supporting folders
- Built subfolders within these based on the positioning doc
- Moved the positioning document itself to "Internal/Positioning\_Documents"
- Asked clarifying questions via multiple-choice prompts about where visual assets should go
When I dumped all my existing client documents in a holding folder and asked it to file them properly, it read each one, understood context, and moved them to the appropriate locations. Not perfectly - I had to correct a few - but with about 90%+ accuracy. To do so myself would have taken hours, and the dread of doing so had prevented me from even trying – until now.
## Hoffman's Coordination Layer in Practice
Hoffman's thread neatly encapsulated exactly what I was seeing: *"The biggest language workload inside any enterprise is the coordination layer: Meetings, notes, docs, action items, status updates, etc."*
That's what my folder structure serves. It's not the intellectual property; it's the scaffolding that lets me find it when I need it.
*"The goal is turning the organisation's memory into something structured and retrievable, so you stop relying on whoever happened to be in the room."*
As somebody with the lousiest of memories, this was and is the potential game-changer – and something I’ve been using a variety of tools to try and achieve for years, with varying degrees of success. How many times have I recreated a positioning doc because I couldn't find the original? How much time goes into "I know we discussed this, but where did we capture it?"
*"AI lives at the workflow level, and the people closest to the work know where the friction actually is."*
Speaking from experience, the friction isn't writing the brief. It's the 15 minutes hunting for the last brief, so you don't duplicate work. It's reorganising folders when a project evolves. It's the mental overhead of "where should this live?"
In other words, *not* the fun stuff.
## The Structure
Wednesday afternoon, I asked Cowork to create a positioning playbook for the new project, similar to what I'd built for Faur but tailored to the new platform's specific needs.
It asked clarifying questions through a clean UI:
Then it created a comprehensive 15-section playbook covering:
- Platform purpose and positioning
- The three-part ecosystem (Learn - Do - Implement)
- Target audiences and their specific needs
- Key differentiators emphasising tested workflows
- Personality and voice guidelines
- Content strategy and quality standards
- Visual identity incorporating Faur's colour palette
This wasn't generic; it was specific to everything I have created previously and was directly informed by this. It had read the entire folder structure, understood how the latest project relates to Applied Comms AI and Faur, and created something genuinely useful.
Thursday, I reorganised my entire Faur directory to properly incorporate all three properties. I asked Cowork to analyse the new structure and update the positioning playbook accordingly. It did. I reviewed and revised for a final version, but all those elements from the first draft were carried through, often pretty much untouched.
## The Shift
This isn't about writing better. It's about removing the structural friction that stops you from writing at all.
How much time do comms professionals spend:
- Hunting for that brief from three months ago
- Recreating positioning docs because the original is buried somewhere or lost in team turnover
- Manually filing meeting notes
- Reorganising folders when a project evolves
- Maintaining the basic hygiene of information architecture
Again, from Reid’s thread: *"Also: Coding agents collapse the cost of analysis, which changes the kind of questions enterprises can afford to ask."*
Replace "coding agents" with "coordination agents." If the time and subsequent cost of maintaining structural organisation collapses, it changes what you can actually build.
When the coordination layer runs itself, you can maintain:
- Three interconnected businesses without structural chaos
- Proper information architecture without dedicated time
- Updated positioning playbooks that evolve with the work
- File systems that make sense even when projects change
## The Assessment
### **What worked brilliantly**
- Understanding context from documents
- Making logical structural decisions
- Creating comprehensive playbooks from scattered information
- Asking clarifying questions before making assumptions
### **What needed human oversight**
- Some files did go to suboptimal locations
- Some positioning nuances required refinement
- Initial folder structures felt excessively complicated. Claude feels a bit too eager to impress at times, but thankfully it does understand the phrase ‘calm your beans’
### **Why keeping backups matters**
This is the first AI that directly changes your actual working infrastructure. That's powerful. That's also why you gate it off and keep copies.
### **The permission gating**
Cowork asks before making changes. It shows you what it's going to do. You approve or deny. This isn't autonomous chaos ([hello Clawdbot](https://www.theregister.com/2026/01/27/clawdbot%5Fmoltbot%5Fsecurity%5Fconcerns/)) - it's collaborative organisation.
## Why This Feels More Significant Than ChatGPT For Comms Pros
When ChatGPT launched, it felt revolutionary because it could write your first draft. But you still had to:
- Open the document
- Find where to save it
- Remember where you saved it
- Organise it within your existing structure
- Update related documents
- Maintain the architecture
Cowork handles the scaffolding. Which means your brain can focus on the thinking that actually requires judgment.
*"Enterprise AI gains compound if you make them shareable. AI lives at the workflow level, and the people closest to the work know where the friction actually is."*
For comms professionals, that friction isn't prompt engineering for content. It's the coordination layer that surrounds every piece of content we create.
## What This Means
Most comms teams are waiting for their organisation to deploy some grand AI strategy. Meanwhile, you can:
- Let Cowork organise your team's shared folders
- Maintain positioning playbooks that evolve with projects
- Create structures that new team members can actually navigate
- Stop losing institutional knowledge in poorly organised drives
Hoffman says the winners will be "companies that build the muscle of day-to-day use early."
For comms professionals, that muscle isn't about better prompts for content generation. It's about letting AI handle coordination so you can focus on the strategic thinking that actually needs human judgment.
- **Next week:** I'll show you what happened when I used Claude Code to build a production website in four days - without being a developer. Spoiler: comms professionals are about to shift from "content creators" to "system designers."
---
*Michael MacLennan is the founder of* [*Faur*](https://faur.site/?ref=appliedcomms.ai)*, a communications consultancy, and* [*Applied Comms AI*](https://appliedcomms.ai/?ref=appliedcomms.ai)*, a platform testing AI tools for communications professionals. He holds an ML/AI certificate from Imperial College London and serves on the board of ScotlandIS.*
### A Complete New Business Pitch & Comms Suite in Under 1 Hour: Testing Claude Skills & Projects
URL: https://www.appliedcomms.ai/claude-skills-projects-business-pitch/
Last updated: 2026-02-25T15:37:05.000Z
**How I built a client pitch from scratch and produced a full campaign toolkit – speaking proposal, presentation, content calendar, outreach emails – in a single session (and doing so in a very meta fashion...). Every asset is shared below.**
---
I receive cold pitch emails every day, and I bet you do too. Most are painfully generic – the same templated approach sent to hundreds of prospects, hoping something sticks. They take seconds to delete because they clearly took seconds to create.
However, what if the economics of tailored outreach could change entirely?
To explore the current capabilities of Anthropic's Claude features – Skills and Projects in particular – I decided to test how they could transform cold business development from a numbers game into something genuinely personalised – and whether I could create a complete, production-ready pitch toolkit within the space of a lunchtime.
- **The target:** Supporting [Faur](https://faur.site/)\* at DataFest 2026, Scotland's premier data and AI festival (27-28 May in Edinburgh), providing a comprehensive initial marketing and communications suite for this potential new client, as well as materials to secure a speaking slot at the event for its founder.
- **The approach:** build everything from scratch in a single session, including the brand guidelines that would inform every asset.
- **The result:** A speaking proposal, presentation deck, content calendar, outreach emails, and an advisory note. All assets shared below, completely unedited from what Claude produced.
- **The required resource:** 90 minutes elapsed time – under 60 minutes of active work, with gaps while Claude processed larger documents. I'm on its $20/month Pro plan, and also occasionally had to wait a couple of times as I'd maxed out its usage limit through using its 'most capable' Opus 4.5 model, with it resetting around 5 hours later.
Faur is my own company, so in essence, I was hypothetically creating this as a cold approach from someone who hadn't spoken to us before. A bit meta, though also a fun way to judge results.
---
## What Are Claude Projects and Skills?
For readers unfamiliar with these features: Claude Projects let you create persistent workspaces where you can upload reference documents, set custom instructions, and maintain context across conversations. Rather than re-explaining your brand, your objectives, or your constraints every time, you brief Claude once and it retains that knowledge.
A "Skill" takes this further – it's essentially a structured set of instructions that tells Claude how to approach specific tasks, and which is stored within the platform for recurring use by Claude itself. For this experiment, I built a brand skill for Faur that specified typography, colour palette, voice principles, and formatting preferences – and did so by asking Claude to create this through looking at [the Faur website](https://faur.site/). This process took a few minutes, then a few more since it hadn't realised the right fonts, which I screengrabbed and located through a free online tool. (Of course, I already knew these, but ignored my own pre-existing knowledge in the interests of the test.) It first created a Word doc, which I then asked to save as a Skill.
The front-loaded investment in creating these instructions promises to pay dividends across every asset produced within the project, and certainly did so in this case.
---
## The Experiment Setup
I set myself a genuine challenge: create a complete pitch toolkit for DataFest 2026 that I could actually provide to Faur alongside a work proposal, and could myself then use. (So thinking about this as not *just* a hypothetical exercise, but real materials for a real opportunity.)
The session included:
- Building Faur's brand Skill in Claude from scratch, and then a complementary Folder for further work (transferring the main conversation with Claude there once ready)
- Researching DataFest (dates, themes, partnership options, The Data Lab's positioning)
- Developing strategy (why Faur should be there, what angle to take)
- Creating the full suite of assets
Hitting the (5-hour) limit
Everything happened in a single conversation. In hindsight, I'd approach future projects differently – creating the brand skill first, then using separate conversations within the project folder for each major deliverable. But testing the single-session approach was part of the experiment.
---
## What Got Produced (With Honest Assessment)
Here's what emerged, with my genuine evaluation of each output's quality:
### 1\. New Business Pitch Overview
A concise strategic pitch setting out why DataFest 2026 is a strong opportunity for Faur, how AI-enabled workflows can be used to create a full, production-ready campaign in minimal time, and why this “show, don’t tell” approach reinforces Faur’s positioning as an implementation-led consultancy.
**Quality: Decent.** This was well structured and produced for what it was (hence receiving a high mark), but felt too much like an internal document – this was on me and my specification! In retrospect, I'd have produced an approach email to the client, tailored to grab their attention, before then providing some of this detail and a link to the rest of the initial 'What We've Created' suite.
- [View the New Business Pitch Overview](https://claude.ai/public/artifacts/88610dc1-d632-4214-bfee-36448582b0ae)
### 2\. Executive Advisory Note
A strategic document outlining why DataFest matters for Faur, the opportunity, risks, and recommended approach. The kind of thinking I'd normally do for a client considering a speaking opportunity.
**Quality: High.** This needed light refinement rather than rewriting. The strategic thinking was sound and the framing matched how I'd approach the analysis.
- [View the Executive Advisory Note](https://docs.google.com/document/d/1OUHwbyicLHUqD5MuB4xyTgwtaUt9AHG-/edit?)
### 3\. Speaking Proposal to Provide to DataFest
A two-page document with session abstract, speaker bio, audience relevance, and technical requirements. Formatted with headers, pull quotes, and Faur branding.
**Quality: High.** Again, refinement territory. The positioning aligned with what I'd have written manually, and the structure followed industry conventions.
- [View the Speaking Proposal (Word doc)](https://docs.google.com/document/d/1QyVYX-Gg0UKZcdYoeU7VjMdI-ERnLgWG/edit?usp=sharing&ouid=107876854741546562988&rtpof=true&sd=true)
### 4\. Presentation Deck
Twelve slides outlining how AI can transform communications workflows, designed to support the speaking proposal.
**Quality: Poor.** Here's where honest assessment matters most. The written content was strong – clear narrative, good structure, appropriate depth. But visually? Version 1 ignored the brand fonts entirely. I requested improvements for Version 2, which was better but still had stylistic problems. Not client-ready without significant design work.
v1 Presentation: Nope!
This matches my broader experience: Claude handles written content impressively well, but visual design remains a clear limitation.
v2 Presentation: Better, but still...
- [View Presentation V1 (PowerPoint)](https://docs.google.com/presentation/d/1mgBBD2YFGP8W0u6YdRNIMwFuoXOQoPsC/edit?usp=sharing&ouid=107876854741546562988&rtpof=true&sd=true)
- [View Presentation V2 (PowerPoint)](https://docs.google.com/presentation/d/1YfxMmr7OOm%5F-vq7AHkd1YDC8T9z5m49D/edit?usp=sharing&ouid=107876854741546562988&rtpof=true&sd=true)
### 5\. Content Calendar
Twelve weeks of social media content across LinkedIn, Instagram, and Threads – covering pre-event thought leadership, event-day coverage, and post-event follow-up. Both company channel and personal profile versions.
**Quality: High.** Comprehensive and immediately usable. The content themes aligned with Faur's positioning and the platform-specific formatting was correct.
- [View the Content Calendar](https://claude.ai/public/artifacts/4eab162c-cc75-40be-849e-ec0420135ce0)
### 6\. Outreach Email
Ready to send to the DataFest team, referencing specific details about the event and making a clear case for the speaking proposal.
**Quality: High.** Personalised, professional, appropriate length. Not a generic template.
- [View the Outreach Email](https://claude.ai/public/artifacts/a6640a91-8c1a-4034-8f28-ec706ab6dae7)
### 7\. This Article! (First early draft)
Yes, asking Claude to produce an article documenting the process was part of the experiment. What you're reading is heavily edited but still based on the original draft, which you can find linked below for comparison. It was pretty solid, but there were elements – specifically Skills – which I wished to emphasise more, as well as providing more comprehensive analysis of performance and materials that the read could access and learn from.
- [View the Original Article Draft](https://claude.ai/public/artifacts/e81f0eb3-20ed-4040-b893-cc16b23ee720)
(Additional note: looking at the screengrab above, think I might actually prefer this intro, but rather than revert I'll let you be the judge...)
### Additional Assets
[View the Full Session Summary](https://claude.ai/public/artifacts/a7bdd98c-9f71-4fe0-bb65-bbed568c76c9) – Claude's own documentation of our conversation, useful for understanding the workflow.
---
## Practical Implications
Some thoughts around what this ability and level of performance means:
- **For business development:** Comprehensively tailored outreach becomes economically viable, being achievable within minutes. Instead of choosing between generic volume and time-intensive personalisation, you can produce genuinely customised materials at scale. The cold emails clogging your inbox exist because personalisation was too expensive. That calculation has changed.
- **For campaign development:** The same project-based approach works for any client or initiative. Build the brand Skill once, create a dedicated project folder, then produce consistent assets across multiple deliverables. What previously took days might take hours.
- **For quality control:** AI doesn't mean AI-approved. Every output would need review and refinement before proper client presentation. Written content required refinement; visual design required significant work. But refinement is faster than creation. The 90-95% quality on written assets means starting from a strong foundation rather than a blank page.
- **For workflow design:** My single-conversation approach worked, but wasn't optimal. Better practice: create your brand Skill, then use separate conversations for each major deliverable within the project folder. You maintain context while keeping individual tasks focused.
---
## Try This Yourself
The approach is replicable:
1. **Create a Claude Project** for your client or initiative
2. **Build a brand Skill** covering voice, visual identity, key messages, and formatting preferences. Ideally use existing brand guidelines, but if unavailable you can provide existing assets (such as a website, graphics, other content) to Claude and use that as a foundational base
3. **Upload reference documents** – add existing content, brand guidelines, relevant research, etc as files to your project folder
4. **Set project instructions** outlining your objectives and deliverable specifications
5. **Work systematically** – strategy first, then assets, reviewing and refining as you go
The time invested in setup pays dividends across all outputs. And unlike traditional pitch work, the assets have multiple uses: the content calendar works regardless of whether DataFest accepts the proposal; the presentation adapts for other speaking opportunities; the article serves Applied Comms AI independently. The pitch becomes inventory, not overhead.
---
## Next Steps
I'll be reviewing the developed materials – as a punter rather than as a creator – in advance of attending [DataFest on 27-28 May](https://thedatalab.com/datafest/). If you're there, let's connect!
And if you're interested in implementing similar approaches for your organisation, I'm opening up to new consultation work in Q1 2026\. Whether that's building custom project workflows, running AI implementation workshops, or auditing your current communications processes, [get in touch](mailto:michael@faur.site).
---
*Michael MacLennan is the founder of* [*Faur*](https://faur.site)*, a communications consultancy, and* [*Applied Comms AI*](https://appliedcomms.ai)*, a platform testing AI tools for communications professionals. He holds an ML/AI certificate from Imperial College London and serves on the board of ScotlandIS.*
### I Built a Voice-Enabled AI Strategy Architect That Generates Bespoke Campaign Dashboards – Here's What Made It Possible
URL: https://www.appliedcomms.ai/voice-enabled-ai-strategy-architect-campaign-dashboard-google-gemini/
Last updated: 2026-02-25T15:38:20.000Z
*How agentic AI turned Google’s Gemini 3 Pro hackathon experiment into a functional and innovative strategy tool for under-resourced charities.*
---
Every charity deserves a professional communications strategy. The problem is that developing one typically takes (at least!) 15–20 hours of skilled work – time and expertise that smaller non-profits simply don't have. So they skip strategy entirely, jump straight to tactics, and campaigns underperform.
In deciding to take past in Google’s [Vibe Code with Gemini 3 Pro](https://www.kaggle.com/competitions/vibe-code-with-gemini-3-pro-in-ai-studio) hackathon last week, I set out to test how far this latest 'ground-breaking' AI model could actually close this gap. Not to just generate text, but deliver something approaching agency-quality output, complete with the strategic rigour, sector knowledge, and practical constraints that real charity communications require. (In the past, [I’ve founded a charity](https://www.thirdsectorawards.com/finalists/michael-maclennan-u0005) as well as taken on senior interim leadership roles and served as a trustee.)
The result was **Charity Voice**: a voice-enabled AI strategy architect that transforms spoken campaign briefs into comprehensive communications strategies, delivered as bespoke interactive dashboards.
In 12.5 hours (10 for development, 2.5 for a recorded video walk-through and entry submission), I built something I genuinely couldn't have created before – and which wouldn’t have been possible for users even a year ago without at least a five-figure budget.
Here's what I learned.
---
v0.1 - Nope, pretty sure this output won't help users...
## **The Background & Technical Detail**
The competition ['Google DeepMind - Vibe Code with Gemini 3 Pro in AI Studio'](https://www.kaggle.com/competitions/gemini-3) was hosted by Kaggle, with Google DeepMind offering $10,000 in Gemini API credits to 50 winners.
Given that the December 2025 release of the Gemini 3 Pro model had caused ChatGPT[ to declare a code red](https://www.wsj.com/tech/ai/openais-altman-declares-code-red-to-improve-chatgpt-as-google-threatens-ai-lead-7faf5ea6?gaa%5Fat=eafs&gaa%5Fn=AWEtsqc6yAXK-kwXnd660hGedCFesqm82twE6oySfMIMHECt%5FeeicLKm2iGVZJotohs%3D&gaa%5Fts=69426df8&gaa%5Fsig=Z2eAD4BDnbjUCmaMF97oEdwtiY6MBScL7B1SMb-d3nH2bsxP98hi8d-30AntAiBhF4a7QQ8kUPHIucuDBC2fAg%3D%3D) – such was its reputed capabilities, it felt like all the excuse I needed to test it within Google AI Studio ("a free, web-based platform for quickly prototyping, building, and experimenting with Google's generative AI models like Gemini").
For the most part I used the free preview available within the studio, until what felt like a mini-crisis near the close for entry submissions, when I ran out of credits. Thankfully, I was then able to pay for use of the full version to get me over the finish line – this experiment therefore costing me a grand total of $0.06.
In terms of creation, I mainly worked with Claude's new Opus 4.5 model to ideate and produce the plan, aiming for something achievable and submittable within 10 hours that would also push boundaries and create something beyond what is possible with the models provided by other rivals. (A slighly ironic use of Claude in that respect.) About halfway through, I also used Gemini 3 Pro's own chat mode separately, as this was better for understanding and building on its capabilities and limits – bouncing ideas and using both models as project partners. More on this later.
---
The breakthrough: When I realised the potential for customisable dashboards (note that not all elements are useful, yet...)
## **The Breakthrough: From Text Output to Tailored Command Centres**
By version 0.2 of the prototype (around hour 2/3), I had something that worked – but it wasn't distinctive. The AI was generating text-based strategies, which is exactly what every other generative AI model has been doing for two years. Useful, but not transformative.
The breakthrough came when I stopped thinking about AI as a writing tool – which would build a whole strategy document based upon my own custom knowledge base – and started thinking about it as a building tool.
What if the output wasn't a document at all? What if it was a fully interactive, visually dynamic dashboard — a "Campaign Command Centre" tailored to each charity's specific campaign?
This shifted everything. When a user completes a strategy session with Charity Voice, they don't receive a Word document or PDF. They get a bespoke HTML dashboard that *can* include elements such as their strategic context, target audiences, a 6-week action plan, key messages, strategic rationale, etc – all presented in an interactive format they can actually use.
For the demo scenario (a Manchester food bank's Christmas appeal), the Command Centre included campaign tracking, audience breakdowns, and timeline visualisation. Other test runs generated fundraising trackers that allowed users to enter donations and see progress against targets. Each dashboard was unique, based solely on what the user had indicated they needed and the supplementary files they had attached.
The combination of voice input (speak your brief naturally, like talking to a colleague) and tailored visual output (receive a working tool, not just text) is something beyond what I've seen other generative AI alternatives offer right now. This isn't just "AI helps me write" – it's "AI helps me build tools."
---
v0.3: Each campaign dashboard is distinct and addresses the challenge presented
## **What I Learned About Agentic Development Using Gemini 3 Pro**
The hackathon forced clarity on what makes agentic development different from running prompts through a standard chat interface.
### **Agentic Code Generation Changes What's Possible**
The standout capability wasn't text generation: it was Gemini 3 Pro's ability to plan multi-step solutions, generate code, and produce working outputs. The model doesn't just respond to prompts; it architects solutions (for this reason, at point I named the app as Charity AI Strategy Architect, though that ultimately felt a bit clunky).
For communicators, this represents a fundamental shift. We've spent two years learning to use AI as a writing assistant. Agentic systems such as that offered by Google Gemini 3 Pro/Gemini AI Studio open up a different possibility: AI as a tool-builder that can create bespoke applications tailored to specific needs.
Annotation: This feature was particularly useful when making changes during the latter stages
### **Voice-First Design Removes Friction**
The second breakthrough was genuinely usable voice input. Users can describe their campaign naturally – speaking rather than typing – and the AI transcribes and understands intent, even with imperfect phrasing.
This matters for accessibility. Many charity staff are juggling multiple roles; asking them to write detailed briefs adds friction. Speaking a brief while walking between meetings removes that barrier entirely. At the same time, the multimodel input allows users to type and/or upload relevant files.
### **Context Windows Enable Grounded Outputs**
Gemini 3 Pro's 1-million-token context window allowed me to load substantial reference materials to ground the outputs in genuine sector practice rather than generic advice. I included an anonymised version of a real communications strategy I'd developed for a charity client (with all confidential information removed), which helped the AI understand what a professional charity comms strategy actually looks like.
This grounding proved essential. Without it, the outputs felt like plausible AI-generated text. With it, they felt like credible first drafts from someone who understands the sector. The structure changed: sections appeared in the order a real strategist would present them, with a decent level of detail in the right places. The quality of insights improved, too, moving from generic advice you'd find in any marketing textbook to recommendations that reflected how charity communications actually work in practice.
---
## **The Development Process: A Dual-AI Workflow**
One of my key learnings was developing a workflow that used different AI tools for their relative strengths.
I used **Claude Opus 4.5** (Anthropic's latest lauded model for complex work) for strategic thinking and prompt engineering — working through the system prompt architecture, developing the 4500-word prompt that defines Charity Voice's personality and methodology, and planning the overall approach. Claude excels at this kind of structured strategic work.
I used **Gemini 3 Pro** (in a separate window from AI Studio) for capability testing: checking what was actually possible within the platform, understanding limitations, and refining prompts based on real behaviour rather than theoretical capability. AI Studio doesn't have a built-in "chat with the model about the model" mode like some platforms (hello [Replit](https://replit.com/)), so running a parallel conversation proved invaluable.
This dual approach prevented the common trap of designing something elegant that the target platform can't actually execute. **I'd recommend it for anyone building AI tools: use one model for thinking, another for reality-checking.**
### **The Build Timeline**
The build went through four versions across ten hours:
**Version 0.1 (Hours 0–3):** The process of arriving at something usable. Basic prototype: worked, but felt like standard chatbot interaction. This phase was about getting the foundations right: voice input working, basic strategy output generating, the conversation flow making sense.
**Version 0.2 (Hours 3–5):** Attempting to shape something genuinely helpful. I added more structured outputs and refined the conversational flow, but the changes started to feel incremental. We were getting stuck in a rut: each iteration made small improvements, but the outputs still felt like every other AI tool generating text strategies. I could sense we were polishing something that wasn't distinctive enough to matter.
**Version 0.3 (Hours 5–7):** The breakthrough. Rather than continuing to iterate, I stepped back and did additional research into Gemini 3 Pro's capabilities: specifically, what it could build rather than just write. This led to a complete restart with a new prompt architecture. The realisation that the output needed to be visually dynamic and engaging, not just well-written, changed everything. Deciding on a bespoke webpage format, and recognising that Gemini 3 Pro could actually generate working HTML dashboards, was the turning point. This version introduced the Campaign Command Centre concept and voice-first design.
**Version 0.4 (Hours 7–8):** Refined conversational flow, added quick-reply buttons, implemented the "Critical Friend" reality check.
v0.4: Using the annotation mode for rooting out smaller issues
### **The Annotation Mode Discovery**
One unexpected discovery: Google AI Studio's annotation feature lets you draw directly on the preview and mark changes visually. By hour 7, I was circling interface issues and writing notes directly on screenshots rather than describing problems in text.
This became my primary method for refinement. I learned that the AI was significantly better at implementing single, visually marked changes than at processing batched text feedback. When I submitted several lines of feedback at once, not all changes would be incorporated — some would be missed or partially implemented. But when I circled a specific element and wrote "make this button larger" or "move this section above the timeline," the implementation was precise.
For visual refinement in AI Studio, annotation mode is faster and more reliable than any prompt-based approach. If you're doing similar work, I'd recommend switching to visual annotation as soon as you're past the core functionality stage.
Hours 8–10 were iterative refinement: fixing bugs one at a time, each marked visually on a fresh screenshot.
The final 2.5 hours went to video production and submission. If you're entering similar hackathons: budget this time properly. A working prototype that you can't demonstrate is worth less than a slightly rougher prototype with a compelling video. My own video efforts were less than my best as a former filmmaker, but I uploaded a quickly edited effort before I entirely lost the will to carry on...
---
## **The App Features That Matter**
### **The Critical Friend Reality Check**
If a charity's goals don't match their resources, Charity Voice challenges them constructively. Claim you'll reach 100,000 people with one part-time staff member and no budget? The AI pushes back before generating a strategy, ensuring recommendations are actually achievable.
This emerged from my real-world charity work. I've seen organisations set impossibly ambitious targets, receive strategies that assume infinite capacity, then fail to execute — or worse, burn out the founders/small teams trying. Many new charities fail or flounder because unrealistic expectations meet limited resources, leaving a profound gap between aspiration and achievement. A tool that just says "yes" to everything isn't actually helpful. It needs to be honest about what's achievable.
### **Grounded Sector Research**
Before producing the campaign command centre, Charity Voice uses Gemini's Search integration to ground strategies in the current UK charity sector context – pulling in relevant data from sources such as the Trussell Trust on food poverty and the current Charity Commission guidance. This isn't hallucinated expertise; it's researched context.
### **Consultative Information Gathering**
Rather than demanding a comprehensive brief upfront, Charity Voice asks smart follow-up questions one at a time – reflecting how a real consultant would conduct a discovery session. Users can respond via voice, typing, or quick-reply buttons, reducing friction at every step. This makes the process more organic and easier for those who may not have thought through every step before hand, with an aim to increase friendliness and completion rate, while providing extra unexpected insight on the way.
### Quick-Reply Buttons for Faster Navigation
Alongside voice input, I built in clickable text prompts throughout the interface – campaign type selection, budget ranges, quick responses to common questions. Users can tap rather than type or speak when they prefer.
This is something I've wanted to implement in charity contexts for almost a decade. When I was working with service users during my charity leadership roles, I saw how much friction-free text input creates for people who are time-poor, stressed, or less confident with technology. Pre-set options that capture common responses make tools genuinely accessible rather than theoretically accessible.
Being able to implement this in a few hours of hackathon development – something that would previously have required significant custom development – felt like a small validation of how agentic AI is changing what's buildable.
---
## **The Honest Limitations**
### **Platform Constraints**
**No Persistence (Yet):** The main weakness is platform-inherent: Google AI Studio doesn't support user accounts. Users can't log in, save their work, and return to iterate over time.
There's a workaround that I quickly implemented: users can download their Campaign Command Centre as an HTML file, then re-upload it to continue refining. It works, but it's clunky. For a fully featured production tool, I'd export the code and build proper persistence elsewhere.
### **Prototype-Stage Limitations**
**Right-Sized Outputs:** Because the primary output is an interactive dashboard rather than a long-form document, strategies run to approximately 1,000–2,000 words rather than the 10,000+ words of a traditional agency strategy. But this may actually be appropriate for the audience. A charity with one part-time comms person doesn't need – and can't act on – a 50-page strategy document. They need something focused, actionable, and implementable. The dashboard format forces prioritisation rather than comprehensiveness for its own sake.
**Refinement Needed:** This is a hackathon prototype, not a polished product. The strategic quality was solid but would need refinement for production release – it felt like there was some room to strengthen and feel less generic. Testing across fundraising, awareness, and advocacy campaigns showed the approach works; the execution needs further iteration.
---
## **What This Means for Communications Practice**
I went into this experiment sceptical about whether AI could replicate strategic thinking – and realise it in an interactive format – rather than just generating plausible-sounding text. My conclusion: we're closer than I expected, with important caveats.
**What worked well:**
- Structured frameworks (stakeholder mapping, channel strategy, risk registers) that AI can reliably produce when given clear templates
- Sector-specific context, when properly grounded with research tools and real reference materials
- Voice input that removes barriers for time-pressed practitioners
- The "consultative" pattern of asking follow-up questions rather than demanding comprehensive briefs upfront
- Interactive outputs that go beyond documents to become working tools – this feels like the real evolution in capability
**What still needs human oversight:**
- Judgment calls about organisational culture and politics that don't appear in any brief
- The intrinsic sense of whether recommendations will actually land with specific stakeholders
- Strategic prioritisation that accounts for unstated constraints and sensitivities
- Quality assurance before anything goes to stakeholders or leadership
- Nuance around timing, competitive positioning, and sector dynamics
At the moment, the Charity Voice prototype is a powerful first-draft generator and strategic thinking partner, *not* a replacement for experienced practitioners. But for charities that currently have *no* strategic input – and there are thousands of them – it could function as a significant step up from nothing.
For comms leaders in larger organisations, a more interesting application might be using tools like this to democratise strategy capability across teams. If every campaign manager can generate a credible first-draft strategy in minutes, senior strategists can focus on refinement and oversight rather than starting from blank pages.
---
## **What's Next**
This experiment has shaped my thinking about where AI-assisted communications is heading. In the new year, I'll be exploring AI agent-driven communications more deeply, with a focus on practical application and quick accessibility through the tools most communicators already have: ChatGPT, Claude, and Gemini.
The goal isn't to chase the latest capabilities for their own sake, but to identify what's genuinely useful for working practitioners right now.
---
The submitted prototype: Available for testing
## **Try It Yourself**
The submission is live and publicly accessible:
- [**AI Studio App**](https://aistudio.google.com/apps/drive/1cunP6vTfwonCHs%5F6m6HL4-QteQHKJ9JX) — try Charity Voice directly
- [**Demo Video**](https://www.youtube.com/watch?v=yDI5c0kL59g) — 2-minute walkthrough of the Manchester food bank scenario
Results from the hackathon judging aren't in yet. But regardless of outcome, this experiment validated something important: voice-enabled, agentic AI tools for communications strategy aren't theoretical anymore. They're buildable today.
---
*I help communications teams move from AI experimentation to operational value. Through* [*Faur*](https://faur.site/)*, I offer workflow audits, implementation consulting, and capability-building workshops—grounded in the same hands-on approach you see in this content.*
*If you're exploring how AI could transform your communications practice, drop me a line at* [*michael@faur.site*](mailto:michael@faur.site)*.*
### 2026 Communications Trends: Sharper Voices Thrive as Tools Get Smarter
URL: https://www.appliedcomms.ai/2026-communications-trends/
Last updated: 2026-02-25T15:37:47.000Z
## **Faur Sight × Applied Comms AI**
*A joint edition for anyone navigating both the human and technical sides of modern communications.*
This month’s piece is shared across [Faur Sight](https://faur.site/insights/2026-communications-trends) and [Applied Comms AI](https://www.appliedcomms.ai/2b794b5ae088809a948ac8520813c55a?pvs=25) because the trends shaping 2026 cut across both worlds: the strategic, stakeholder-facing work of communicators, and the practical, hands-on reality of AI-driven workflows. This joint edition keeps the analysis in one place — clear, simple, and useful wherever you’re reading it.
So, without further ado…
---
## **The 2026 Landscape: Calm Heads Required**
If 2025 was the year everyone rushed to “add AI” to their workflow – with decidedly mixed results – 2026 is the moment where organisations realise that technology alone doesn’t deliver clarity, trust or credibility.
Audiences are overwhelmed.
AI-generated content is everywhere.
The teams who stand out now are the ones who combine technical competence with unmistakably human judgment and practical experience – the thread that underpins every trend below.
Let’s dig in.
---
## **1\. AI Quality Becomes a Communications Issue, Not a Tech Issue**
2025 revealed a simple truth: the problem isn’t “AI,” it’s **unreviewed AI**.
We saw brands, government bodies, and media organisations publish content containing:
- Minor hallucinations
- Inaccurate details
- Generic voice ([and ‘em dash’ controversies](https://www.theringer.com/2025/08/20/pop-culture/em-dash-use-ai-artificial-intelligence-chatgpt-google-gemini))
- Tone mismatches
- Contradictory statements
- Unintentional bias or outdated information
All because AI was treated as an autopilot, rather than a writing partner.
### **2025 example**
**Air Canada’s chatbot incident** became a global case study. Their website’s AI system confidently generated an incorrect refund policy, and [the courts held the airline liable](https://www.bbc.com/travel/article/20240222-air-canada-chatbot-misinformation-what-travellers-should-know). This wasn’t a tech failure; it was a communications failure rooted in poor oversight.
### **What to do now**
Produce and regularly refine AI editorial standards, ensuring to review workflows and introduce solid governance protocols – nothing like a solid approval process! – before any damage is done.
---
## **2\. Comms Moves Closer to Product: Explainability Becomes a Must-Have**
In 2026, communicators will need to understand the mechanics behind their products, systems, and AI tools, since audiences now ask *how* things work, not just *what* they do. (I particularly enjoyed [Elena Verna’s piece around this](https://www.elenaverna.com/p/brand-a-product-job-now) for the Elena's Growth Scoop newsletter.)
Regulators, journalists, customers, and employees all expect clarity around:
- Data use
- Algorithmic decisions
- AI-assisted processes
- Safety mitigations
- Risk trade-offs
If you can’t explain it, people are likely to assume the worst.
### **2025 example**
Whenever **Google has launched new Gemini updates**, the company has faced intense scrutiny over how the model handles safety interventions – forcing Google’s comms, policy, and engineering teams to jointly publish explainers, breakdowns, and blog posts clarifying how the system works. It became a textbook case of comms and product joining forces, and a strengthened approach [has been evident in the recent Gemini 3 launch announcement](https://www.wired.com/story/google-launches-gemini-3-ai-bubble-search/).
### **What to do now**
Create “explainability summaries” for all major launches, especially where AI is involved. If your team can’t summarise it clearly, your audience definitely can’t interpret it confidently.
---
## **3\. The Return of the Homepage (Owned Channels Matter, Yet Again)**
AI-overviews reshaped search behaviours in 2025\. Social platforms splintered further (a subject from [my 2025 trends piece](https://faur.site/insights/2025-business-communications-trends) which was borne out, and then some). Algorithmic visibility became inconsistent at best, nonexistent at worst.
Organisations realised the platforms they don’t control can’t be relied upon for reach or accuracy.
Expect a renaissance in:
- Email newsletters
- Homepage updates
- Content hubs
- Owned community spaces
- Subscription ecosystems
### **2025 example**
**The New York Times** and several major publishers [doubled down on direct subscription products](https://www.nytimes.com/2025/11/05/business/media/new-york-times-earnings.html) after reporting their lowest year of social referral traffic in a decade. Some outlets saw *over 60% drops* from platforms like Facebook and X – pushing them to re-centre their homepages and apps as the primary audience gateways. The increasing prevelance of Google’s AI Overview in its search results [was reported to be causing ‘devastating drops’](https://www.theguardian.com/technology/2025/jul/24/ai-summaries-causing-devastating-drop-in-online-news-audiences-study-finds) back in the summer – and this direction feels only one way.
### **What to do now**
Build owned channels like media assets: ensure that you combine consistent rhythm, clear value, strong UX, and direct relationships.
---
## **4\. Executive Visibility Goes Multimodal (a.k.a Your Face Matters)**
Leaders can’t communicate through text alone anymore.
Stakeholders expect faces, voices, and spontaneous human presence.
LinkedIn continues to be key for professional audiences, but the increasing amount of obviously AI-written content – we’ve all noticed it – has made audiences sceptical and more likely to scroll on past, unless you provide them with a near-instant reason not to.
The best-performing executive communication in 2025 leaned into authentic multimodal formats:
- Short video explainers
- 30–90 second audio updates
- Livestream Q&As
- CEO “voice notes”
- Behind-the-scenes walkthroughs
It wasn’t about polish – it was about being real.
### **2025 example**
For its June 2025 WWDC rollout, Apple supplemented the keynote and press release with short video walkthroughs from senior product leads, 30-60s “feature explainer” reels, and conversational interviews across mainstream media. The multimodal mix helped demystify what could otherwise have been an opaque technical launch, and made the company and its C-suite team feel more approachable – iterating on a technique which Steve Jobs famously pioneered.
### **What to do now**
Create low-friction setups for leaders – simple workflows, clear prompts, and formats that feel natural. This can be achieved through the simple, practical use of AI agents within the major providers, while ensuring an approval process with an extra pair of eyes before anything goes live.
---
## **5\. Community Becomes a Strategic Asset, Not a Side Quest**
Communities are increasingly where trust lives.
In 2025, brands that didn’t engage them honestly felt the consequences.
Communities aren’t audiences – they’re networks with their own norms, influencers, moderators, and internal logic. They require careful stewardship rather than broadcast tactics, especially since they can live within a complex network of differing platforms, ranging from familiar social platforms and Reddit through to private customised community platforms and WhatsApp channels.
### **2025 example**
In 2025, [local newsletters with subscriber models](https://www.niemanlab.org/2025/10/are-these-local-newsletters-local-news-and-does-it-matter/) increasingly became the primary place residents go for timely, verified updates. Civic organisations and even local authorities began sharing information *through newsletter writers* because audiences trusted them more than official accounts or social platforms.
Here in the UK, you can see the same shift in action with [the excellent Edinburgh Minute](https://www.edinburghminute.com/), [whose founder I interviewed last year](https://faur.site/insights/edinburgh-minute) (and to which I’m a paid member myself, due to its concisely informative format). Its rapid growth and unusually high levels of trust show how community-led publishing increasingly shapes information flows – and why communicators need to treat newsletter creators and community stewards as core stakeholders, not optional extras.
### **What to do now**
Treat community spaces the way you treat stakeholder groups: thoughtfully resourced, clearly governed, and genuinely engaged.
---
## **6\. Human Insight Becomes the Differentiator Again**
Here’s the thread running beneath every trend:
**AI can generate content, but it cannot judge what matters:**
- It can’t weigh political nuance…
- It can’t sense reputational risk…
- It can’t decide what to say at a tense town-hall meeting…
- It can’t feel the emotional temperature of a community...
- It can’t choose the one sentence that shifts a room…
…well, not most of the time, anyway, and certainly not reliably.
2026 is the year when communicators should loudly and proudly articulate that their value isn’t in *producing* content – it’s in making sense of the world.
### **2025 example**
In 2025, for the first time, [social media overtook TV](https://www.niemanlab.org/2025/06/for-the-first-time-social-media-overtakes-tv-as-americans-top-news-source/) as Americans’ top news source – a trend reflected across the world. Multiple studies show that audiences now rely more on human-led explainers than on institutional channels, for everything from product recommendations to understanding complex issues. People increasingly favour creators whose personality, consistency, and judgment they recognise – because they feel they have a relationship with them. It’s a powerful reminder that clarity, trust, and human voice are still what cut through the noise, and that organisations can replicate them by putting real people – and real judgment – at the heart of their communication.
### **What to do now**
Invest in taste, narrative skill, stakeholder sense, and the human skills that AI can’t replicate. That’s your moat, both in the immediate and longer term.
---
## **Final Thoughts: 2026 Rewards the Thoughtful**
These trends aren’t about chasing shiny tools and attention-grabbing gimmicks: they’re about building systems, workflows, and habits that deliver clarity, trust, and credibility:
- AI is now foundational.
- Owned channels are essential.
- Personally delivered communication is expected.
- Communities hold power.
- Human judgment binds it all together.
---
## Need help with Digital & Communications?
If you're wrestling with AI workflows, digital strategy, executive visibility, or just need someone to help cut through the noise, let's talk.
I work with organisations who want communications that actually work: clear strategy, human-centred execution, and practical AI implementation where it makes sense.
Check out [who we are](https://faur.site/about), [what we do](https://faur.site/offers), or just say hello: [info@faur.site](mailto:info@faur.site)
You can also find us on [LinkedIn](https://www.notion.so/faursight/link), [Instagram](https://www.notion.so/faursight/link), and [Threads](https://www.notion.so/faursight/link).
### The AI Architect Building the Future of Communications at FleishmanHillard
URL: https://www.appliedcomms.ai/interview-fleishmanhillard-joyce-higgins/
Last updated: 2025-10-31T11:39:59.000Z
### **While many comms leaders are still asking 'should we use AI?', Joyce Higgins is already several steps ahead, building custom AI environments that transform how entire teams work.**
When Joyce Higgins joined [FleishmanHillard](https://fleishmanhillard.com/), in May 2024, she kept mentioning that AI would fundamentally change how we search for and consume information. "Everyone was like, 'Yeah, I guess that's something we should have on our radar,'" she recalls. Fast forward to today, and as Head of AI Strategy, she's leading transformation across one of the world's largest communications agencies, building custom AI environments that are revolutionising how teams operate.
Here's what makes Higgins different from the countless AI evangelists flooding our LinkedIn feeds: she's not interested in talking about AI, she's interested in building with it. Most importantly, she's interested in teaching others to build with it too.
"You can talk about AI for the next 12 months, but if you haven't built anything, you're not an AI specialist," she states. It's this practical approach – honed through years of building complex digital systems – that's driving real transformation at FleishmanHillard.
## The Foundation: From Complex Logic to AI Architecture
Understanding Higgins' approach to AI requires understanding where she came from. Her first major project at [Digitas Health](https://www.digitashealth.com/) was a $10 million TV-driven copay registration card system which integrated CRM channels through complex backend business rules. "We did all of the business rules and marketing operations behind the online and offline mailers," she explains. "To this day, it was one of the most complex marketing projects I've worked on."
This early immersion in data-heavy, logic-driven systems set a pattern. When social media exploded across healthcare clients, Higgins didn't just adopt it – she built the operational procedures from scratch. When she moved to Real Chemistry, she didn't just join their digital team – she built their entire digital activation capability for the London office, including paid media, digital strategy, integrated planning, and measurement teams.
> Reality Check #1: "AI expertise isn't about being technically gifted – it's about understanding systems and workflows."
By 2020, she was already working with conversational AI at [Real Chemistry](https://www.realchemistry.com/), building tools specifically trained on handling healthcare-related questions for niche audience segments. "We had a tool that was specifically trained on handling healthcare-related questions for very niche audience segments, such as medical science liaisons," she notes. This wasn't theoretical exploration: it was a practical application for real clients with real needs.
So at the start of 2025, Higgins was ready to take on her new role FleishmanHillard as their Head of AI Strategy. Not just because she could talk about large language models or explain how neural networks function, but because she understood how to build systems that actually work.
## The Framework: Building the AI Comms Toolbox
The centrepiece of Higgins' approach is deceptively simple: the AI Comms Toolbox. But don't let the straightforward name fool you. This is a sophisticated framework that transforms how teams adopt and scale AI capabilities.
"Essentially, it is a suite of agents, the instructions for those agents, and guidance on how to set up the knowledge documents for your average communications account," Higgins explains. Teams receive the complete package, which includes process documentation, testing criteria, agent instructions, knowledge document examples, and six weeks of hands-on support.
The framework operates on a crucial principle that many organisations miss: everyone needs to become AI-capable, but not everyone needs to become an AI expert. "Our strategy is that everyone has to be an AI specialist," says Higgins.
However, this isn't about turning PR professionals into programmers. It's about democratising AI capability across the entire organisation.
> Reality Check #2: "People get nervous. They say, 'I don't understand how to manipulate data.' But I explain: you can write digital content, and that's all it is. When you're writing instructions for an agent, it's just describing to someone how something should be achieved."
The toolbox requires teams to implement at least three types of agents:
1. An insights agent (for analysis and understanding)
2. A content development agent (for creation)
3. A review agent (for quality control and compliance)
This isn't arbitrary. Higgins has discovered that most people think of AI purely for content generation, missing its powerful applications in content review and analysis. "We have agents that support any kind of code of practice," she notes. "Not everyone appreciates how valuable it is for content review."
The six-week sprint structure provides just enough support without creating dependency. Week by week, teams progress from nervous newcomers to confident builders. "What we've noticed is it becomes essentially training wheels," Higgins observes. "Something that seemed really scary to them before now seems completely achievable."
## The Build: From Theory to Working Systems
Here's where Higgins' approach diverges from typical AI adoption strategies. Rather than rolling out generic tools and hoping teams find uses for them, each account team builds customised solutions for their specific needs.
FleishmanHillard utilises Omnicom's proprietary AI infrastructure, which offers enterprise-level security and enables the development of sophisticated knowledge bases. "It gives us a lot of ability to build out custom knowledge bases for our clients, and then the agents off of the back of it in these really secure sandboxes," Higgins explains. Teams can upload hundreds of documents to create deep institutional knowledge about their clients.
But the technical infrastructure is just the starting point. The real innovation is in how teams learn to think about AI applications. Take virtual audiences, for instance. Instead of relying solely on publicly available data from tools like GWI or Talkwalker, teams are building virtual representations of specific communities using focus group research and qualitative data.
"We found that going out and getting focus group research and data and bringing it back and building virtual audiences to make sure that you're really creating messages that resonate with different audiences and communities is the best way to go about it," Higgins explains. This isn't AI for AI's sake – it's AI solving real communications challenges.
> *Reality Check #3: "Everyone who is using AI is seeing that it's not saving a huge amount of time. If folks are saying it's saving a huge amount of time, they're probably lifting copy directly from the platform and using it as is, which is not an effective way of using Gen AI."*
The value isn't in the time saved, but in the capability gained. Teams can bring more insights to the table, communicate with greater precision, and allocate more time to strategic planning rather than tactical execution. "There's a lot more time for planning and having those strategic conversations, rather than the actual writing of the words," Higgins notes.
## The Debugging: Learning from What Doesn't Work
One of the most refreshing aspects of Higgins' approach is her willingness to discuss what doesn't work. In an industry often obsessed with polished case studies, she's forthright about the failures and challenges.
"There are two bad habits of communicators that I always tell the team to please break," she says. The first is the tendency to talk about AI rather than build with it. The second – and perhaps more pernicious – is the need for perfectionism.
"Communicators want precedent. They want to know where we've done it before, and they don't want to show the client anything that isn't perfect," Higgins observes. "But the problem is that you have to show stuff that's not perfect because you don't want to go too far down the line in a build."
This perfectionism paralysis is particularly acute in communications, where professionals are trained to present polished, final versions of everything. But AI development requires an experimental mindset. "You're not going to innovate if there's precedent," Higgins points out.
The debugging process has revealed other crucial insights. When teams first encounter AI tools, many become nervous about the technical requirements. "People were a little bit nervous. It felt a bit scared. It's AI, it's different. I didn't have to do digital before, so why do I have to do this type of digital?"
The breakthrough in breaking down resistance came when Higgins reframed the conversation: "This isn't a digital tool, this is a comms tool."
## The Scale: Extension Packs and Exponential Growth
Once teams master the basic toolbox, something remarkable happens: they start building their own solutions. "By the end of our six-week sprint, they're able to start building their own custom agents," Higgins notes. This is where the real transformation occurs – when AI adoption shifts from top-down implementation to bottom-up innovation.
To support this growth, FleishmanHillard has developed what Higgins calls "extension packs" – pre-built modules for specific communications needs:
- Crisis communications agents
- Editorial content systems
- Thought leadership report builders
- IP and intellectual property reviews
- Media relations tools
- Measurement and analytics agents
- Brand voice and tone checkers
- Regulatory compliance reviewers
- Multi-language adaptation tools
- Community sentiment analysers
"When success is really realised is when we notice that an account is so confident in their use of AI that they start to say, 'Okay, for this particular work stream, I need this extension pack, because AI is now just the way that we work,'" Higgins explains.
The modular nature allows teams to mix and match components based on their specific needs. A healthcare client might combine regulatory compliance with community sentiment analysis. A consumer brand might focus on brand voice with multi-language adaptation.
> Reality Check #4: "Now is a really forgivable time to be comfortable with experimenting. Say to your stakeholders, 'I was trying something new. It wasn't perfect, but we have a lot of key learnings.'"
## Sign up for Applied Comms AI
Practical AI insights for forward-thinking comms leaders.
Subscribe
Email sent! Check your inbox to complete your signup.
No spam. Unsubscribe anytime.
## The Ethics Architecture: Building Responsible AI
Higgins doesn't shy away from the ethical implications of AI in communications. In fact, she sees communicators as having a unique responsibility in shaping how these tools develop.
"Words and ideas and narratives are the new data," she emphasises. "We don't need to know an additional language to produce any kind of data that's going to train these large language models. Communicators have a bit of a responsibility to make sure that we round out some of the biases that we do see."
This isn't just philosophical musing. FleishmanHillard now has formal AI agreements with clients that detail logging procedures, paper trails, copyright considerations, and QA processes for identifying harm or bias. Teams are trained not just in using AI, but in using it responsibly.
The virtual audiences work exemplifies this approach. Rather than accepting the limitations of publicly available data, teams actively seek out diverse perspectives and underrepresented voices to build more complete representations of their audiences.
## The Commercial Reality: Making the Business Case
How do you justify AI investment to sceptical leadership or clients? Higgins has a refreshingly practical approach: show them the money.
"What I have found works the best is bringing a commercial assessment," she says. The team measures specific impacts: speedier content delivery times, junior team members becoming more autonomous, senior team members having more time for strategic planning, and faster onboarding for new team members.
"When you actually do the commercial assessment of how much time is either saved or redistributed to a higher value opportunity, there's very little conversation after that," Higgins notes. "The leadership team always says, 'How do we get this done?'"
But she's careful to set realistic expectations. This isn't about replacing people or dramatically cutting costs. It's about elevating capabilities and redistributing effort to higher-value activities.
## The Future Architecture: What's Next
Looking ahead, Higgins is most excited about two developments: collaborative AI environments and predictive analytics for communications.
"The ability to collaborate with clients in a sort of shared environment" is the next frontier, she explains. Imagine combining Omnicom's PR-first tools and data sets with a pharmaceutical company's real-world evidence and audience data. "The ability to match those two types of data sets to create this really comprehensive view of what their audiences are and then build the AI from there is going to be really, really exciting."
Predictive analytics represents another leap forward. "From a comms perspective, I think it's an area where there's more opportunity, especially when we're thinking about how we want to change narratives for our clients, being able to react and adapt to the market and understand how we need to shape the narrative."
But perhaps the most profound change is already here: the expectation of personalisation. "If I go to ChatGPT and it remembers that I have a dog and I like hiking when I go on holiday and I'm travelling with my partner, I may never go read a travel blog post again," Higgins points out.
This has massive implications for communications. "We're so used to creating static, long-form content. We really need to think about these new opportunities to create really personalised experiences for our audiences."
## **Building Your Own Blueprint**
What can communications leaders learn from Higgins' architectural approach to AI transformation?
- **Start with systems, not tools:** The AI Comms Toolbox works because it's a complete system – processes, training, support, and iteration. Don't just throw AI tools at your team and hope for the best.
- **Make everyone a builder, not a user:** The goal isn't to have a few AI experts and many passive users. It's to enable everyone to build and experiment within their expertise area.
- **Embrace imperfection:** "You have to show stuff that's not perfect," Higgins insists. The teams that progress fastest are those that prototype, test, fail, and iterate quickly.
- **Measure what matters:** Don't focus on time saved – focus on value created. Where is effort being redistributed? What new capabilities have you gained?
- **Think modular:** The extension pack approach means you can start simple and add sophistication over time. You don't need to transform everything at once.
- **Own the responsibility:** As communicators, we're not just using AI – we're training it with our words, ideas, and narratives. That comes with an obligation to do it thoughtfully and inclusively.
Joyce isn't interested in the AI hype cycle. She's not waiting for perfect tools or clear precedents. She's building the future of communications, one agent at a time, one team at a time, one experiment at a time.
"Everyone's going to expect to be able to have a dialogue with some sort of interface and get a super personalised experience back to them," she predicts. "We're going to be much more clever about the way we do comms moving forward."
The question isn't whether AI will transform communications – it's whether you'll be one of the architects of that transformation, or just a passenger along for the ride.
As Higgins puts it: "You can talk about AI for the next 12 months, or you can build something."
Which will you choose?
---
- [***Connect with Joyce on LinkedIn.***](https://www.linkedin.com/in/joycehiggins/)
- *This interview is part of Applied Comms AI's "Leaders & Learning" series. Sign up for regular conversations with communications leaders navigating the AI revolution, alongside our hands-on comms AI experiments and investigations.*
- *Applied Comms AI is powered by*[ *Faur*](https://faur.site/?ref=appliedcomms.ai)*.*
### AI in Crisis Comms: Help or Hindrance?
URL: https://www.appliedcomms.ai/crisis-comms-ai/
Last updated: 2026-02-25T16:08:45.000Z
Even in today’s comically unpredictable landscape, crisis communications remains the ultimate high-stakes industry challenge: one where minutes can determine reputation survival and precise wording requires a level of surgical precision.
A swathe of swanky-sounding AI tools have entered the marketing, promising revolutionary speed and scale. However, they come with significant risks: such as the notorious hallucinations that introduce falsehoods, over-alerting that creates noise, or sensitive data handling that creates compliance nightmares.
For the discerning comms professional, we wanted to take a closer look at each stage of crisis comms, to see where AI may produce tangible practical benefits – and also when and how to remain cautious and make considered choices.
What follows is a playbook for using AI as a potential force multiplier while maintaining human judgment at critical decision points, one which can guide us at this time and point the way to future opportunities.
To help us navigate this landscape, we're joined by Amanda Coleman, one of the UK's leading crisis communication experts.
With over two decades managing communications for Greater Manchester Police – including some of the UK's most challenging incidents – Amanda literally wrote the book on crisis communications (three of them, in fact). Now Director of Amanda Coleman Communication Ltd, she advises organisations internationally on crisis preparedness and response. As a Chartered PR practitioner and Fellow of both CIPR and PRCA, Amanda brings hard-won frontline experience to the AI conversation.
---
## **Step 1: Early Signals & Monitoring**
Machine learning has powered social listening for years, through tools such as[ Brandwatch](http://brandwatch.com/),[ Talkwalker](https://www.talkwalker.com/), and[ Meltwater](https://www.meltwater.com/): AI arguably just makes it faster, cheaper, and available to teams without enterprise budgets.
**Where AI helps**
- **Volume processing:** Monitor 10,000+ mentions daily vs 100s manually - catch issues hiding in the noise
- **Pattern recognition:** Spots unusual comment velocity (50 complaints in an hour vs normal 5) before humans notice
- **Multi-language alerts:** Real-time translation means that viral TikTok in Portuguese doesn't blindside you tomorrow
**Where it hinders**
- False positives, sarcasm/irony blind spots
- Missing private/low-visibility channels (CS tickets, WhatsApp, internal chat)
**Amanda's view:** “Using AI for social media monitoring should be in place and is almost an entry level way of getting comfortable with the technology and what it can do. But it can also be subject to hallucinations which can derail the approach by putting the focus in the wrong place.
“Where it gets really interesting is predictive analytics – AI can spot a problem when it's starting to emerge, but as we get more sophisticated, we should be able to have alerts that the conditions exist for a problem or crisis to develop. This will allow swift action and possible mitigation, which elevates the risk management processes.”
- **Try this for free:** Set up[ Google Alerts](https://www.google.co.uk/alerts) *and*[ Talkwalker Alerts](https://www.talkwalker.com/alerts) for: "\[Your brand\]" OR "\[CEO name\]" AND (crisis OR complaint OR boycott OR outage)"
---
## **Step 2: Situation Assessment**
The promise here is tantalising: turn information chaos into clear briefings in seconds, not hours. But here, hallucinations and contradictions can be especially troublesome.
**Where AI helps**
- **Source synthesis:** Combine 20 news articles, 50 tweets, and 5 internal emails into one coherent timeline
- **Fact extraction:** Pull key data points (numbers affected, locations, times) from walls of text within seconds
- **Contradiction flagging:** Can highlight discrepancies, such as when your PR says "resolved" but customer service says "ongoing"
**Where it hinders**
- Papering over contradictions; hallucinating links; missing one-off critical facts
**Amanda's view:** “AI doesn't always access the latest information, and needs to be brought in house to avoid any confidentiality concerns.
“It could be used to assess your messaging and whether it's aligned to brand values and is within any legal and regulatory boundaries. But be aware of the challenge of biases creeping into the systems. AI, like people, is not infallible.”
*Experiment with this prompt (public info only):*
*You're a crisis communications manager conducting rapid situation assessment.*
*Your role is to create a clear, factual briefing for senior leadership.*
*Summarise the following sources into a 200-word SITREP with these sections:*
1. *CONFIRMED FACTS - Include source for each*
2. *UNCONFIRMED REPORTS - Label as "UNCONFIRMED:"*
3. *STAKEHOLDERS AFFECTED - List all groups impacted*
4. *SPREAD/SCALE - Geographic reach, platforms affected, volume of complaints*
5. *IMMEDIATE RISKS - What could escalate in next 2-4 hours*
*Additional instructions:*
- *Flag any contradictions between sources with "CONTRADICTION:"*
- *Use bullet points for clarity*
- *Include timestamps where available*
- *Highlight any regulatory/legal mentions*
- *Note if any media outlets are covering this*
*Sources to analyse:*
*\[Paste PUBLIC links/excerpts only – no personal data\]*
- **Compliance note:** If personal data is involved, *don’t* paste it into public tools. Use enterprise LLMs with a DPA, or anonymise. Check out ‘[UK ICO guidance: AI fairness/transparency and when to run a DPIA’.](https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/artificial-intelligence/guidance-on-ai-and-data-protection/?utm%5Fsource=chatgpt.com)
---
## **Step 3: Strategy & Message Development**
This is where AI walks a tightrope: fast drafting is invaluable, but one tone-deaf statement can torpedo trust permanently.
**Where AI helps**
- **Speed drafting:** Generate 5 holding statement options in 2 minutes vs 30 minutes manually
- **Tone calibration:** Test if your apology sounds defensive ("we regret any inconvenience") vs genuine
- **Completeness checking:** Flags missing elements (no timeline for fix, no contact for questions)
**Where it hinders**
- Robotic phrasing; unintended legal admissions; cultural tone-deafness
**Amanda's view:** “Studies recently show 69 per cent of people can't tell what is real and what is not. This is the issue for approaches and messaging. We have to make sure that it is clear it is human and has empathy. Having the right prompt will provide a better outcome but without it can be wide of what is needed.
“Also, messaging isn't a process as in if you do X and Y then Z will happen. This makes it vital to put the people into the message and strategy development. There's a lack of understanding of people's requirements – who are you speaking to and how can you frame the messages accordingly?”
---
## **Step 4: Stakeholder Engagement**
Different audiences, same truth: AI can help you translate without contradicting yourself, *if* you're careful about the data you feed it.
**Where AI helps**
- **Message versioning:** Turn one core statement into eight variants (investors, staff, customers, media) maintaining consistency
- **Concern prediction:** Maps likely questions per group (staff: "are jobs safe?", investors: "financial impact?")
- **Channel matching:** Suggests format per audience (bullets for internal email, paragraph for press)
**Where it hinders**
- Privacy/compliance risk if you upload personal data; over-segmentation causing inconsistency
**Safe use:** Provide *roles/job titles only* (no names/emails). Ask for an influence × concerns × channel × cadence grid and a one-page engagement plan. (If you *must* process personal data, do it inside your CRM or an enterprise LLM instance; see ICO.) ([ICO](https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/artificial-intelligence/guidance-on-ai-and-data-protection/?utm%5Fsource=chatgpt.com))
**Amanda's view:** “Have this data and information available as part of crisis planning. Don't wait until something happens and then turn to AI. It is only as good as the data it accesses. Internal comms staff will always know more about the employees than AI can. Can it understand changes in human behaviour that are identified in some of the recent research on crisis messaging? It often will default to what is known rather than identify the changes.”
---
## **Step 5: Execution & Channel Management**
The nightmare scenario: your LinkedIn says one thing, an internal employee engagement email says another, and someone screenshots both.
**Where AI helps**
- **Platform optimisation:** Adapts your 500-word statement for internal comms, Instagram carousels, LinkedIn posts
- **Timing coordination:** Calculates optimal posting sequence across time zones and platform peak times
- **Link/tag QC:** Checks all handles are correct, links work, hashtags aren't hijacked
**Where it hinders**
- Auto-visuals that feel glib; over-automation can lead to generic, cold, and contradictory copy
**Amanda's view:** "Your approach will be scrutinised – remember American Airlines' statement reused by Air India. If there is a lack of detail to shape the statement and actions then it may not be appropriate to the situation or could become like wallpaper because all statements say the same. Although I would argue that a lot of crisis communication has already gone that way with people writing statements!"
---
## **Step 6: Monitoring & Feedback Loop**
Once your response lands, you need to know if it's working. AI can process the firehose of feedback, but can it understand what it means?
**Where AI helps**
- **Response velocity:** Analyse 1,000 comments in 5 minutes vs 5 hours manually
- **Theme clustering:** Groups similar complaints ("can't log in", "lost access", "account locked" = authentication issue)
- **Escalation flagging:** Identifies high-risk responses (legal threats, media interest, influencer anger)
**Where it hinders**
- Sentiment ≠ truth (irony/sarcasm); volume can drown out influence
**Amanda's view:** “Ultimately, you have to know who you are speaking to, what they think of you before the crisis, and then you can see what impact it has had and whether the communication has worked. Often, we don't have the right data and information to review our effectiveness."
---
## **Step 7: Review & Learning**
After the dust settles, exhausted teams need to capture lessons before memory fades. AI never forgets, but can it truly learn?
**Where AI helps**
- **Meeting efficiency:** 90-minute debrief becomes 20-page transcript with 1-page summary
- **Action extraction:** Pulls out all commitments ("Sarah will update the crisis manual by Friday")
- **Pattern analysis:** Can quickly compare this crisis to previous ones: what issues keep recurring?
**Where it hinders**
- Over-simplifies organisational dynamics; misses emotional/cultural context
**Amanda's view:** “AI can analyse the feedback and data from people during the crisis. However, it may miss the nuances that people gather from a debrief. It needs to include more context.
“Also, can it really take you from a point made on the minutes through to the actions that are needed without understanding the history of the organisation and the crisis being faced?"
---
## **Verdict**
Working hypothesis: from current practical use, it feels as though AI can do approximately 30% of the workload brilliantly (monitoring, clustering, drafting options, triage). However, the remaining 70% stays human: judgement, empathy, trade-offs, politics, and accountability – elements where nuance and authenticity are crucial.
Amanda agrees with this assessment but adds a crucial caveat: "The key is preparation. Use AI now to understand its capabilities and limitations. Build your prompts library when you're calm. Test on low-stakes issues. Because when a real crisis hits, you don't want to be experimenting – you want proven tools and workflows ready to deploy."
---
## **Your Crisis-AI Starter Pack**
**Set up now:**
1. **Talkwalker Alerts + Google Alerts** for brand/execs/products + risk terms. ([talkwalker.com](https://www.talkwalker.com/alerts?utm%5Fsource=chatgpt.com),[ Google](https://www.google.com/alerts?utm%5Fsource=chatgpt.com))
2. **One drafting LLM** (Claude or ChatGPT) + a tiny library of approved prompts.
3. [**Otter.ai**](http://otter.ai) **or Fireflies** for debrief capture → summary → action items. ([help.otter.ai](https://help.otter.ai/hc/en-us/articles/4425393298327-Otter-Notetaker-Overview?utm%5Fsource=chatgpt.com),[ ](https://fireflies.ai/?utm%5Fsource=chatgpt.com)[Fireflies.ai](http://fireflies.ai))
4. **5 prompts to save:** SITREP (Step 2) • Statement red-team (Step 3) • Stakeholder grid (Step 4) • Comment triage (Step 6) • Debrief summary (Step 7)
**Quarterly fire-drill (90 mins):** Also, pick a real but minor complaint to conduct a simulation. Run all seven steps above. Time everything. Note what broke. Fix it before you need it for real.
---
## **Quick GDPR/ICO Checklist (UK)**
✅ Never paste personal data into public AI tools
✅ If processing personal data, use enterprise tools with a DPA (or anonymise)
✅ Run a[ **DPIA**](https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/accountability-and-governance/data-protection-impact-assessments-dpias/) for any new AI crisis workflow
✅ Keep an audit trail: who used what, when, for what purpose
✅ “It was AI” isn’t a defence — *you* are accountable ([ICO](https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/artificial-intelligence/guidance-on-ai-and-data-protection/?utm%5Fsource=chatgpt.com))
---
I hope you’ve found some of the above useful. Do let me know your thoughts, and it would be great to get your feedback on use cases: what are you testing for crisis preparedness? What worked (or backfired)? Reply with your experiments: we’ll feature the best lessons in the next issue.
- For comprehensive crisis communication expertise, Amanda Coleman's books and consultancy services, visit [amandacolemancomms.co.uk](http://amandacolemancomms.co.uk).
### From Watson to Lovable: How Flipside's MD is Navigating the AI Revolution
URL: https://www.appliedcomms.ai/interview-flipside-barney-evison/
Last updated: 2025-08-12T00:00:04.000Z
*Welcome to the first in our "Leaders & Learning" series – regular conversations with communications leaders who are actually experimenting with AI, not just talking about it. We're cutting through the hype to share what's working, what isn't, and what's being learned along the way.*
*Our inaugural interview features Barney Evison, Managing Director at Flipside, the innovative digital agency that's part of Weber Shandwick, and is a big part of* [*its new ‘Weber I/O’ AI offering*](https://lbbonline.com/news/Weber-Shandwick-Launches-Weber-IO-in-EMEA)*. Barney has been experimenting with AI since 2017, long before ChatGPT made everyone an overnight expert. From early natural language processing experiments to today's no-code revolution, he's engaged in enough experiments to know what actually works.*
*What struck me most about our conversation wasn't just the practical insights – though there are plenty – but Barney's refreshing honesty about the gap between AI promise and reality. This feels like* a *– if not* the *– vital component for building long-term trust with clients and stakeholders when it comes to pushing the boundaries with the underlying needs and goals firmly front and centre.*
*– Michael MacLennan*
*(Conversation recorded 30 July 2025)*
---
## **Top 5 Takeaways**
1. **Show, don't tell**: Getting AI-sceptical team members to experience that "magical moment" with tools like *Lovable* beats any amount of evangelising
2. **Failed experiments are gold**: That automated video generator that didn't work? It taught Flipside exactly where AI's current limits lie
3. **"Agentic" is the new "synergy"**: Most AI buzzwords are meaningless – focus on gradual transformation, not revolutionary promises
4. **Local AI models could change everything**: Smaller, secure, company-specific models might be the real enterprise opportunity
5. **The commoditisation paradox**: No-code tools that threaten agency models also open doors to clients who couldn't afford custom development
---
## **The Interview**
When Barney Evison, Managing Director at Flipside, joined the agency in 2016, he thought he was escaping the "wider trappings and bureaucracy" of large agencies like his previous employer, Edelman. Six months later, the independent digital shop was acquired by Weber Shandwick. "I went into it being like, 'Oh great, I'm doing this small independent agency,'" he laughs. "But actually, it's been an amazing experience."
That adaptability has served him well. Flipside has been experimenting with AI since 2017 – using IBM Watson's APIs when most of us were still figuring out chatbots. Today, he's navigating a very different landscape: where clients don't just want AI capabilities, they want "AI" as the answer to everything.
"Everyone had that feeling when ChatGPT launched in November '22," Evison recalls. "That moment of like, 'this is so much better than I thought.' I asked it to write a press release in the style of Edgar Allan Poe, and the results felt was amazing."
But unlike many who stopped at amazement, Evison and his team have spent the past few years turning that wonder into practical applications – and learning some hard truths along the way.
## **From Watson to ChatGPT: The Evolution**
Flipside's AI journey began with natural language processing (NLP) back in 2017-18, using IBM's Watson for sentiment analysis and chatbots. "Watson was probably the best thing on the market at the time that was commercialised," Evison explains. It was relatively sophisticated – API connections, custom solutions, social media monitoring – but it all happened "under the hood."
Then came the ChatGPT moment that changed everything. Suddenly, AI wasn't just a background technology; it became the main event. "Now clients are actually coming and asking for AI as a thing in itself," he notes. The shift has been seismic: from AI as invisible infrastructure to AI as the star of the show.
Today, Flipside operates on two fronts: helping companies transform their workflows for an AI world, and building AI-powered solutions from scratch. They're working with everyone from FTSE companies needing enterprise solutions to startups whose entire business model is AI-powered.
The speed of change in the startup space particularly astounds him. "*Lovable* recently broke records in terms of the speed at which it reached $100 million in annual revenue," he says. These aren't gradual success stories – they're explosions.
**→ What this means for your team:** The shift from "AI under the hood" to "AI as the product" requires different skills and positioning. Start preparing your team for both technical implementation and strategic AI conversations.
## **The Reality of AI Implementation**
So what's actually working at Flipside? The unsexy (but highly important) stuff, mostly. An internal Statement of Work (SOW) generator that transforms proposals into commercial documents complete with legal boilerplate. User story generators for the design team. Image generation for rapid prototyping.
"Creating SOWs, we have an internal generator which we now use," Evison explains. "We create a proposal or strategy, then very quickly turn that into a commercial document that already has boilerplate legal copy and all the stuff around invoicing and exclusions."
It's not glamorous, but it works. These tools tackle the mundane, repetitive tasks that eat hours but add little value. The tools that stick share common traits: they're focused, practical, and augment rather than replace human expertise. *Waldo* for brand research. *Perplexity* for strategy work. Standard stuff, but deployed thoughtfully.
Then there's the surprise success – and existential challenge – of no-code development. Working with a small business client recently, one of Flipside's developers used no-code tools to build a functional web application at a fraction of the traditional cost. "In full transparency with the client," Evison emphasises.
"It's an interesting dilemma we're confronted with," he reflects, "the commodification of some of the things we actually offer." But rather than resist, they're leaning in – using these tools to serve clients who previously couldn't afford custom development, while focusing their premium services on complex challenges that still require human expertise.
**→ What this means for your team:** Start with tedious but time-consuming tasks. The ROI is clearer, the risk is lower, and you'll build confidence for bigger experiments.
## **What Doesn't Work (And Why That's Valuable)**
Not every experiment succeeds. Evison's automated case study video generator stands as a perfect example of AI's current limitations. The vision was compelling: feed in written case studies, get back polished videos. The reality? "After going back and forth on it, doing prototypes for a while, we realised it was better to just literally brief a video editor."
The creative and production-focused AI tools – video content, avatars, automated content generation – consistently disappoint. "They're moving really fast, but it's still quite fledgling," he says. "They'll get better, but they're not quite there."
This matters because of a phenomenon Evison observed in recent research: developers using AI tools often feel faster without actually being faster. "People using these tools sometimes feel like they're being more efficient, but the actual time it takes isn't faster. They're actually developing sometimes even more slowly."
It's an insight that cuts through the hype: just because something feels innovative doesn't mean it's effective. Flipside's failed experiments have taught them to measure actual outcomes, not perceived progress.
**→ What this means for your team:** Track time saved, not just tasks completed. The perception of efficiency can be deceiving – measure real results.
## **Leading Through the Scepticism**
"Designers, developers, copywriters – people with a craft tend to be more resistant," Evison observes. It makes sense: if you've spent years honing a skill, a tool that promises to replicate it in seconds feels threatening.
His solution? Let them experience the magic themselves. "I definitely had that addictive moment trying *Lovable*," he recalls. "Wanting to just spend ages building stuff in it." When someone who can't code suddenly creates a functional website, it changes their perspective on what's possible.
But forcing adoption doesn't work. Instead, Flipside has appointed AI champions in each vertical, letting specialists lead adoption in their own domains. "I don't go to a designer and say, 'you need to use this tool for all your UI designs,'" Evison explains. "I get them to bring their own views to the table."
They've also created clear guardrails. Teams want to experiment but need boundaries. "Don't put client data into X tool" – that sort of thing. It's about creating a safe space for experimentation while managing risk.
**→ What this means for your team:** Let your biggest sceptics become your champions. Give them the tools, the guardrails, and the freedom to find their own "aha" moment.
## **The Myths and Misunderstandings**
"People use 'agentic' like people use 'societal' – they just use it to sound smart."
The quote, from analyst Benedict Evans at a recent event, perfectly captures Evison's frustration with AI buzzwords. "Agentic" has become the latest meaningless term, thrown around to make mundane automation sound revolutionary.
"Being properly AI-powered where you've got it doing things autonomously, making decisions without human oversight – that's a massive change," Evison clarifies. "You can't just build one or two automated workflows and be like, 'we're agentic.'"
Fundamental transformation requires rewiring how companies operate, not just bolting on a few tools. This reality check extends to how agencies think about AI investment. "Now clients are like, 'we need to be doing something on AI,'" he notes. The risk isn't not investing – it's investing badly. Building complex "agentic workflows" when your data isn't ready. Buying massive licences when half your team won't use them.
**→ What this means for your team:** Focus on gradual transformation over revolutionary promises. Small, consistent improvements beat grand gestures every time.
## **Looking Forward: Opportunities and Challenges**
Evison's most excited about local models – smaller, open-source AI that companies can run securely on their own hardware. "You can set up locally hosted models quite easily," he explains. Imagine a marketing team with their own custom AI model, trained on their brand voice, running entirely within their firewall. "Really safe, and the quality can actually be better."
Crisis communications is another frontier. "AI can help you move much more quickly," he suggests, imagining systems that sense-check responses in real-time during emergencies. Not publishing automatically – that's still too risky – but accelerating human decision-making when speed matters most. (Weber I/O [recently launched Radius](https://webershandwick.co.uk/news/weber-shandwick-launches-weber-i-o-in-emea-elevating-data-technology-and-ai-capabilities), an AI-powered issues and crisis solution which uses proprietary algorithms to guide users through a crisis response, doing so in full compliance with their protocols.)
His advice for other agency leaders? "Start experimenting now. Put tools in the hands of the team... give them parameters... maintain that dialogue."
**→ What this means for your team:** The next wave isn't about bigger models but smarter, more focused ones. Start thinking about what a custom AI for your specific needs might look like.
## **The Speed of Change**
"The speed of change at the moment seems much faster than any previous technological wave I've been through," Evison reflects. Social, mobile, cloud – he's seen them all. But AI is different.
That may be why his advice is so straightforward: just start. Not with grand strategies or transformation frameworks, but with experimentation. Give teams tools. Let them play. Learn from failures. Build from successes.
"It's a fascinating time to work in a digital agency," he says. Coming from someone who's weathered acquisitions, technological shifts, and the occasional Edgar Allan Poe press release, that optimism feels earned.
The future Evison sees isn't one where AI replaces agencies, but where smart agencies use AI to do things they never could before. Where "agentic" hopefully goes the way of "synergy" – remembered only as a cautionary tale about buzzword inflation.
Until then, Flipside will keep experimenting, failing, learning, and occasionally experiencing those magical moments when the technology actually delivers on its promise.
- *Barney Evison is Managing Director at Flipside, part of Weber Shandwick, and you can follow him* [*over on LinkedIn.*](https://www.linkedin.com/in/barneyevison/)
- *This interview is part of Applied Comms AI's "Leaders & Learning" series. Sign up to our newsletter for regular conversations with communications leaders navigating the AI revolution, alongside our hands-on comms AI experiments and investigations.*
- *Applied Comms AI is powered by* [*Faur*](https://faur.site/?ref=appliedcomms.ai)*.*
### Building Smarter Pitches and Smoother Approvals: How AI Can Fix PR Bottlenecks
URL: https://www.appliedcomms.ai/newsletter-2025-07-30/
Last updated: 2026-02-25T16:09:25.000Z
Hi all – and a special thanks to those who have been sharing with friends and colleagues, it's been fantastic to see the subscriber count climbing! For this latest newsletter, I've been building actual AI tools, crafting prompts that generate brilliant media pitch subject lines, and discovering how simple folder organisation can transform your AI workflow.
Often fun and occasionally baffling, these experiments have been an insight into how AI can help for daily comms workflows, and with the news that OpenAI is [getting set to launch GPT-5 in August](https://www.theverge.com/notepad-microsoft-newsletter/712950/openai-gpt-5-model-release-date-notepad), it feels like the capabilities will soon be increasing. With that in mind, there seems no better time to get stuck in…
---
## 1\. Building in Public: The "AI Boss Approval Detector"
*Creating functional AI tools without coding: complete with the mistakes, breakthroughs, and "oh, that's why it's broken" moments.*
**What I Built:** The [Faur Stakeholder Lens](https://comms-persona-preview.lovable.app/?ref=appliedcomms.ai) – an AI tool that analyses your content through different stakeholder perspectives before you hit send. Think of it as a digital focus group that's always available.
**Why?** Because we've all sent something that landed badly with a key stakeholder. Hours of crafting, endless tweaks, and you're still blindsided by someone's reaction. What if you could test your message through their eyes first?
**How?** Using Lovable (an AI-powered app builder), Claude Opus 4, and Supabase (an easily integrated database). The hypothesis: AI can simulate different stakeholder perspectives well enough to catch those "should've seen it coming" moments.
### **What Happened**
*Hour 1: The Reality Check*
- Started with Claude-generated prompts. Immediate error. Welcome to no-coding.
- Basic UI working within 30 minutes, complete with persona dropdowns.
- Discovered Lovable's "Implement this plan" feature – it outlines your build, then executes at a click. Revelation for non-coders.
*Hour 2: The "Everything is Fine" Bug*
- Created user log-ins for tailored personas, hit Claude usage limits, switched to ChatGPT.
- Major discovery: wrote a deliberately terrible press release full of jargon and buried leads. The app gave it 95%.
- The reality: The app wasn't actually analysing anything! Everything defaulted to a generic score because things weren’t right under the hood.
*Hour 3: The Breakthrough*
- Fixed the issue (which was related to parsing). Suddenly awful content scored 10-20%. Success!
- Final sprint: Threw seven improvements at Lovable simultaneously. All implemented within minutes. Added sample content, copy functionality, keyboard shortcuts, visual indicators.
**The Reality Check:** For a final test, the Employee Rep persona instantly flagged missing staff impact messaging; Legal caught consultation requirements I'd missed. Five minutes versus potential hours of stakeholder review cycles before getting to the same stage.
**Try It Out:**
[Full breakdown: Building An AI-Powered Comms Studio](https://www.appliedcomms.ai/app-building/ai-stakeholder-comms-review-tool/)
---
## 2\. Prompt Perfection: Media Pitch Subject Lines That Actually Work
*Testing refined AI prompts for your own comms use.*
**Use Case:** Every PR professional's nightmare: crafting the perfect pitch, then agonising over the subject line that determines if it even gets read.
**How We Refined:** Applied the Q&A Strategy (forcing AI to ask clarifying questions first), enhanced role assignment with specific domain expertise, and added structured output with risk assessment.
### **The Prompt**
\`You are an experienced PR professional specialising in media relations with 15+ years of crafting successful pitches. Your expertise includes understanding journalist psychology, news values, and inbox management patterns.
I need you to optimise email subject lines for media pitches using a systematic approach.
First, ask me these 5 strategic questions:
1. What's the core news angle of your story?
2. Which journalist/publication are you targeting and what's their typical beat?
3. What's the most surprising or counterintuitive element of your story?
4. Is there a timely hook?
5. Do you have any exclusive angles or data to offer?
Based on my answers, generate 5 optimised subject lines that:
- Are 6-10 words maximum (50-60 characters)
- Front-load the most newsworthy element
- Include specific numbers/data when relevant
- Avoid PR clichés
- Match the journalist's demonstrated interests
For each subject line, provide:
1. The subject line
2. Why it works (25 words)
3. Risk level (Low/Medium/High) for being ignored
4. Best time to send
Format as a table with clear rankings from strongest to weakest.\`
**In Practice:** Instead of generic options, you get targeted suggestions written with the journalist firmly in mind. The risk assessment helps you choose between safe bets and attention-grabbing gambles.
[Full prompt guide: Building the Perfect AI Prompt for Media Pitch Subject Lines](https://www.appliedcomms.ai/prompt-engineering/ai-prompt-media-pitch-subject-line/)
---
## 3\. Workflow Game-Changer: Why Project Folders Matter
*Tests of AI-powered organisation – what's worthy of inclusion in your comms toolkit?*
**Tool Name & Category:** Project Folders - ChatGPT & Claude (Workflow Organisation)
**The Promise:** Transform chaotic AI conversations into organised, accessible workspaces
**The Test:** Used folders across both platforms for client work, experiments, and content creation over several months
**The Good:**
- Instant organisation for multi-client juggling
- Upload files for constant reference (client strategies, brand positioning, stakeholder lists)
- Pick up exactly where you left off on any project
- Create dedicated conversations like "jargon busters" for quick reference
**The Not-So-Good:**
- Limited folder hierarchy (just lists, not folders-within-folders)
- File handling still hit-and-miss beyond Word and PDFs
- No cross-platform sync between ChatGPT and Claude
**Best For:** Anyone working across multiple clients, projects, or experiments who's tired of endless scrolling to find that brilliant conversation from last week
**Pricing:** Free with ChatGPT Plus and Claude Pro subscriptions (each £20 a month in the UK)
[Full blog: How Project Folders Supercharged My AI Comms Workflow](https://www.appliedcomms.ai/tool-testing/project-folders-ai-comms-workflow/)
---
## Comms AI in the Headlines
News and commentary of note this week:
- **Ragan's new Center for AI Strategy** 'puts people at the heart of AI for comms': "CAIS is founded in a belief that people will lead the AI revolution," said Diane Schwartz, CEO of Ragan Communications ([Ragan](https://www.ragan.com/))
- **How the UK Government is pioneering AI for comms use:** 'Government Communications has pioneered Assist, the first general purpose AI tool approved for use across the UK Government' ([Gov.uk press release](https://www.gov.uk/))
- **PR's AI Originality Dilemma:** 'Are PR firms willing to give up their own creativity for the sake of ease?' ([PR Week](https://www.prweek.com/))
- **AI in customer communication: the opportunities and risks SMBs can't ignore:** AI boosts SMB communications, but trust is critical ([Tech Radar](https://www.techradar.com/))
---
## Thanks for Reading!
I appreciate all feedback, and do let me know if there's anything you would like to see in future editions.
**Worth sharing?** Forward this to a colleague trying to make sense of AI.
**New here?** Subscribe to Applied Comms AI for regular insights →
## Sign up for Applied Comms AI
Practical AI insights for forward-thinking comms leaders.
Subscribe
Email sent! Check your inbox to complete your signup.
No spam. Unsubscribe anytime.
*Applied Comms AI is powered by* [*Faur*](https://faur.site/?ref=appliedcomms.ai)*.*
### Speed-Building an AI 'Will My Boss Hate This?' Detector, Using Lovable
URL: https://www.appliedcomms.ai/ai-stakeholder-comms-review-tool/
Last updated: 2026-02-25T16:13:02.000Z
From idea to working app for an AI-Powered Communications Studio within the time it takes to watch Oppenheimer - bugs, breakthroughs, and all. Here’s what *you* can build in three hours.
## The Problem We're Trying to Solve
Every comms professional has sent something that landed badly with a key stakeholder: even the battle-hardened ones. Hours of drafting, endless tweaks… and you’re still blindsided by someone’s reaction.
So, what if you could *test* and *refine* your message before hitting send? Imagine running your content past a digital stand-in who’s tuned to the same tastes, quirks, and priorities as your boss, client, or that tricky stakeholder. Someone (or something) who gives honest, fast feedback—every time.
Enter the [Faur Stakeholder Lens](https://comms-persona-preview.lovable.app/): An AI tool that analyses your content through the eyes of different stakeholders before you hit send. Beyond a few generic archetypes, the aim is to build specific, realistic AI “doppelgängers” for your actual recipients.
**Why it matters:** If functional, it will save you face, save time, and slash those painful revision cycles by anticipating feedback before it happens.
**Reality check:** This won’t replace human judgement, but it might catch those “obvious” misses that haunt you at 3am. (Or is that just me?)
## The Plan (And Where It Could Go Wrong)
**Our hypothesis:** AI can simulate different stakeholder perspectives well enough to provide useful pre-flight checks on comms.
**Success looks like:**
- Without coding or using expert technical knowledge, building a functional app with some reasonably complex features. (For example, I would never have the slightest idea of where to start with building a sign-in profile for a website)
- Accurately flagging negative reactions to risky or poorly written content
- Giving genuinely distinct feedback for different personas (e.g. CEO vs Legal vs Employee Rep)
- Suggesting actionable improvements, not just “generic” criticism
**Failure, on the other hand, shall resemble:** Every piece of content gets similar scores regardless of quality, or if the personas all sound the same.
**Technical approach:**
- **Platform:** Lovable (AI-powered app builder)
- **AI model:** Claude Opus 4 (and a bit of OpenAI GPT-4)
- **Additional tools:** Supabase for data (I signed up to this through the recommendation of Lovable, and began using during the course of the experiment!)
- **Key features:** Persona-based analysis, content generation, amendment suggestions
## The Process
### Hour 1: From No-Code Prompts to Prototype
Combining Claude and Lovable to get going quickly
- **15:03**: Started with a Claude-generated prompt for Lovable. Immediate error (“UserTie is not a valid export”). Welcome to no-coding.
**15:15**: Fixed. Basic UI now working, with dropdowns for the basic personas and document types.
- **16:05**: Discovered Lovable’s “Implement this plan” feature. It outlines your build, then executes at a click. This is a revelation for non-coders.
- **16:19**: Sidetracked by OpenAI API setup – namely, how on earth to set up the OpenAI API. Thankfully, it was all relatively straightforward, with Lovable pointing to the right place to do so. After this set up Supabase, which would allow me to provide in, so users can build and save their own custom personas. Felt suspiciously simple to do so.
The helpful text box suggestions were AI-generated without any prompting
- **16:24**: *It’s alive!* First version previewed at [comms-persona-preview.lovable.app](https://comms-persona-preview.lovable.app/) (note: this link now previews and shows the latest version!).
**Lesson:** Even with AI’s help, the boring stuff – API keys, import errors – still eats time.
### Hour 2: Features and Discoveries
**16:32** \- Created a custom Keir Starmer persona using Claude – hey, you can dream big! Then hit my Claude Pro usage limit (as a side note, any other Pro users finding they’re quickly hitting their daily limit?), so switched to ChatGPT to generate a test press release.
**16:45** \- First major bug: custom personas crashed the app. Blank page of death.
**17:19** \- Added content generation feature, so that users could quickly get inspiration rather than staring at a blank text box. Interface felt cluttered. Asked Lovable to clean it up and add [Faur](https://faur.site/) branding.
**Lesson:** AI tools can build *fast*, but not always *right*. Testing is essential.
Here is where I was after about one and a half hours, when I stopped for the day:
Results for end of day one
### Hour 2.5: The "Everything is Fine" Bug
Here's where it got interesting, as I picked up the app-building on a new day. I purposely wrote a terrible press release - full of jargon, buried lead, zero news value. The app gave it 95%.
**The discovery:** The app wasn't actually analysing anything. OpenAI was returning markdown-formatted JSON, the parsing was failing, so everything defaulted to a generic 75% score.
**The lesson:** A working UI doesn't mean working functionality. Always test with intentionally bad inputs.
### Hour 3: The Final Sprint
Fixed the JSON parsing issue, even though I’m still pretty unsure about what that is. Suddenly lousy content was scoring 10-20%. A success at spotting duds!
**Final 15 minutes** \- Threw seven quick improvements at Lovable, because why the hell not. Having resumed on a different day, I asked Claude what *it* thought could be improved, and decided to go with each of its suggestions:
- Sample content button
- Copy results functionality
- Keyboard shortcuts (Cmd+Enter)
- Visual score indicators with emojis
- Print-friendly styles
- Footer credits
- Welcome modal
Lovable implemented all seven simultaneously – hitting a few errors but debugging itself as it went – and completing this final major task within a few minutes, which felt vaguely miraculous.
**04:52** \- CSS syntax error. Fixed in 35 seconds. **05:06** \- Replaced "Follow-up Questions" with an "Amendments" feature, so the app actually modifies the user’s content based on feedback. **05:10** \- Made amendments optional and smart (changes tone, structure, clarity based on persona).
## First Test: Does It Actually Work?
**Scenario:** Drafting an announcement about an office relocation (a classic comms headache).
- **Traditional approach:** 2–3 hours of rewrites and stakeholder reviews (potentially days depending on how pernickety the stakeholders are); risk of missing key objections.
- **With the tool:** 5 minutes to test across all personas. The Employee Rep persona instantly flagged staff impact messaging gaps; Legal flagged consultation requirements I’d missed.
*Result:* Not perfect, but surprisingly useful for catching obvious missteps *before* the human review stage, so an undoubted time-saver.
## What We've Learned (So Far)
**About AI in comms:**
- AI can convincingly simulate different stakeholder perspectives (at least for first-pass review)
- The best insights come from testing content across *multiple* personas
- Generated “amendments” are often surprisingly actionable
**About building tools:**
- "What's missing?" prompted to AI can generate comprehensive improvement plans
- Visual feedback (scores, emojis) makes abstract analysis concrete
- Big bugs can hide in the gap between “it looks like it works” and “it actually works”
## What I Would Change
**Immediate priorities:**
- **Stakeholder Showdown:** Compare all personas side-by-side.
- **Export reports:** Download feedback and scores.
- **Team sharing:** Share reviews internally.
**Bigger questions this raises:**
- How much should we trust AI to simulate human reactions?
- Could “pre-testing” with AI risk making comms *too* cautious?
- Will it kill authenticity if everyone optimises for “no objections”?
## The Wrap-Up
Let’s be honest: This will never replace the years of nuanced, political stakeholder judgement that experienced comms professionals bring. It won’t catch every subtlety, or predict the totally irrational responses.
But for catching those “should’ve seen it coming” moments? For giving junior team members a safety net? For that late-night or crisis statement, when you can’t get instant stakeholder input? This is genuinely useful.
**The real revelation isn't the tool itself - it's that a comms professional with no coding experience can build something this functional in three hours.** The barriers between "I wish this existed" and "I built this" are crumbling.
**Try it yourself:**
---
**Found this useful?** Subscribe – if you haven’t already – and please forward this to a colleague who may be facing a similar challenge.
*Applied Comms AI is powered by* [*Faur*](https://faur.site/)*. We help organisations navigate the intersection of communications and technology.*
---
### Building the Perfect AI Prompt for Media Pitch Subject Lines
URL: https://www.appliedcomms.ai/ai-prompt-media-pitch-subject-line/
Last updated: 2026-02-25T16:14:32.000Z
You've crafted the perfect media pitch. The angle's sharp, the stats are compelling, and you've tailored it beautifully for that tech journalist who covers AI. Then you spend another 15 minutes agonising over the subject line, knowing that those seven words will determine whether your pitch even gets opened. Sound familiar?
## The Prompt
Here's what we built:
```
You are an experienced PR professional specialising in media relations with 15+ years of crafting successful pitches. Your expertise includes understanding journalist psychology, news values, and inbox management patterns.
I need you to optimise email subject lines for media pitches using a systematic approach.
First, ask me these 5 strategic questions:
1. What's the core news angle of your story? (product launch, research findings, expert commentary, etc.)
2. Which journalist/publication are you targeting and what's their typical beat?
3. What's the most surprising or counterintuitive element of your story?
4. Is there a timely hook (trending topic, upcoming event, news cycle)?
5. Do you have any exclusive angles or data to offer?
Based on my answers, generate 5 optimised subject lines that:
- Are 6-10 words maximum (50-60 characters)
- Front-load the most newsworthy element
- Include specific numbers/data when relevant
- Avoid PR clichés ("excited to announce", "groundbreaking", "revolutionary")
- Match the journalist's demonstrated interests and style
For each subject line, provide:
1. The subject line
2. Why it works (25 words)
3. Risk level (Low/Medium/High) for being ignored
4. Best time to send (based on the angle)
Format as a table with clear rankings from strongest to weakest.
After presenting options, ask if I'd like variations on any specific approach or need subject lines for different journalist segments.
```
**Quick Start Guide:**
1. Copy the prompt above
2. Fill in the bracketed sections
3. Run through your preferred AI tool
4. Review and refine with human judgment
💡 **Pro tip:** Always A/B test your top two subject lines by sending to different journalists at the same publication
## How We Got Here (The Journey)
### The Problem
Every Comms and PR pro knows the brutal truth: journalists receive dozens of pitches daily. (I certainly did two decades ago back at The Herald, and the volume has only increased since then.)
With pressure on output greater than ever, your subject line has roughly two seconds to convince them not to hit delete (and that’s probably generous). Despite that, most of us have at times defaulted to tired formulas that scream "PR pitch" from a mile away.
Traditional approaches are already dead in the water, thanks to overuse. Today's journalists have developed an immunity to phrases like "thrilled to announce" and "industry-leading solution." They're looking for genuine news value, delivered efficiently.
So, how can we make use of AI to provide a systematic and efficient way to craft subject lines that think like journalists, not marketers?
---
### Version 1: The Overly Basic Approach
When (understandably) pressed for time, the temptation is to bash out something simple, such as:
```
Write a compelling subject line for my media pitch about our new AI product launch.
```
The output? Most likely to be a generic, overpromising line that will barely register second glance, and be indistinguishable from the hundreds of other tech pitches flooding journalists' inboxes.
**Key learning:** Without added context, AI is likely to default to the mean – and in this case, that means PR clichés which will instantly turn the target audience away.
---
### The Iteration Process
**Iteration 1: The Q&A Strategy**
The first major improvement came from implementing the [Q&A Strategy](https://reykario.medium.com/4-must-know-ai-prompt-strategies-for-developers-0572e85a0730) \- forcing the AI to ask clarifying questions before generating output. This prevents the AI from making assumptions and ensures we're working with complete information.
- **The tweak:** Added a structured discovery phase requiring the AI to gather specific context about news value, timing, and exclusivity angles.
- **The result:** Instead of generic subject lines, we now get targeted options that speak directly to what journalists care about. The AI asks about exclusive data, timely hooks, and surprising angles - exactly what makes editors take notice.
**Iteration 2: Role Strategy with Domain Expertise** Next, we enhanced the persona beyond just "experienced PR professional" by adding specific domain knowledge and behavioural patterns.
- **The tweak:** Specified 15+ years of media relations experience, understanding of journalist psychology, and inbox management patterns. This creates a more sophisticated "thinker" who understands the nuances of media gatekeeping.
- **The result:** Subject lines that avoid PR clichés and instead mirror the language patterns journalists use themselves. The AI now thinks like someone who's successfully pitched thousands of stories.
**Iteration 3: Structured Output with Risk Assessment** The final cherry on top came through adding systematic evaluation criteria to each suggestion.
- **The tweak:** Required the AI to provide risk levels and timing recommendations for each subject line, forcing deeper consideration of context and strategy.
- **The result:** Not just subject lines, but strategic intelligence about when and how to use them. This transforms a simple generation task into a decision-support tool.
---
### Real-World Applications
From testing so far, this prompt provides great results across various pitch scenarios:
- **For breaking news:** Front-loads urgency and exclusivity, with high-risk/high-reward options for competitive stories where being first matters more than being perfect.
- **For research launches:** Emphasises surprising statistics or counterintuitive findings that challenge conventional wisdom - the kind of angles that make journalists think "my readers need to know this."
- **For expert commentary:** Connects your spokesperson to trending topics within the 24-48 hour news cycle, positioning them as the go-to voice on emerging issues.
- **For feature pitches:** Focuses on human interest angles and broader trend pieces that fit into journalists' forward planning cycles.
As a former subeditor who has spent far too long mulling over single word choice within a headline, the time savings here could be substantial, However, the real value here isn't speed - it's the systematic approach that ensures you're considering all the angles that make journalists pay attention, providing a variety of options and angles customised for your consideration.
---
### The Technical Bits
This prompt makes the most of several techniques from prompt engineering research:
- The **Q&A strategy** prevents premature convergence on generic solutions by forcing information gathering first. This mirrors how experienced PR professionals actually work - understanding the full context before crafting the pitch.
- The **role assignment** goes beyond surface-level expertise to activate specific knowledge about journalist behaviour and news values. This isn't just about writing better - it's about thinking like the person receiving your pitch.
- The **structured output format with risk assessment** adds a strategic layer often missing from creative tasks. It acknowledges that different situations call for different levels of boldness.
For edge cases – like pitching breaking news where speed matters more than optimisation – you can skip the Q&A phase by providing all context upfront. The framework is flexible enough to adapt to your workflow.
---
### Your Turn
Try adapting this prompt for your specific media lists. Consider:
- Do certain journalists respond better to data-driven subject lines versus human interest angles?
- Should you adjust the word count for mobile-first readers who see even less in preview?
- Would benefit/impact-focused lines work better for trade publications versus consumer media?
Track which subject line styles work best for different types of journalists and publications. The prompt is designed to evolve with your learnings, and can of course be adapted and iterated on as you venture forth. (For example, you could within the prompt provide extra detail about a specific journalist, or point to a database containing additional information.)
---
### The Bottom Line
This prompt aims to solve the subject line paralysis that kills pitch momentum. It's particularly valuable for PR teams juggling multiple clients, stories, and deadlines.
Remember: even the best subject line can't save a weak story, but a weak subject line will definitely kill a strong one. This prompt ensures that never happens on your watch.
---
**Found this useful?** Subscribe – if you haven’t already – and please forward this to a colleague who may be facing a similar challenge.
*Applied Comms AI is powered by* [*Faur*](https://faur.site/)*. We help organisations navigate the intersection of communications and technology.*
---
### How Project Folders Supercharged My AI Comms Workflow (And How They’ll Transform Yours Too)
URL: https://www.appliedcomms.ai/project-folders-ai-comms-workflow/
Last updated: 2026-02-25T16:15:37.000Z
When OpenAI [introduced Project Folders to ChatGPT](https://www.theverge.com/2024/12/13/24320800/openai-chatgpt-projects-folders-ai-chats) in December 2024, it was a small but mighty upgrade: instantly transforming how I organised client work, AI experiments, and content drafts. It followed Claude (Anthropic’s AI assistant), which [had already begun](https://www.anthropic.com/news/projects) offering similar folder features earlier that year.
As someone who now regularly bounces between both platforms, I’ve made Project Folders the backbone of my comms workflow.
If you’re still just relying on memory, bookmarks, or endless scrolling, you’re missing out. Here’s how and why folders have become essential for getting real value from generative AI, especially when it comes to communications, consultancy, and creative work.
---
## The Problem: AI Clutter Happens Fast
AI tools encourage experimentation: capturing fleeting ideas, spinning up drafts, refining messaging, and reviewing to further shape your thoughts. Very quickly, however, your workspace can fill up and descend into a jumble of “Untitled” chats, partial drafts, and promising experiments lost to the feed. Important work gets buried. Refinding a client brief or last week’s brainstorm becomes a headache.
For anyone regularly using ChatGPT or Claude – especially if you work across multiple clients or projects – this quickly becomes unsustainable.
---
## Enter Project Folders: Organising & Optimising Your Comms AI workflow
Folders in Claude were the first real solution to AI clutter. Now that ChatGPT offers a similar feature, you can (and absolutely should) create a coherent system across both platforms.
**What does this mean in practice?**
- For every new project, client, or experiment, I spin up a folder in either ChatGPT and Claude – sometimes both, if I want to get alternate responses and choose the best.
- I then create project instructions: a concise description that can include the overall purpose, how I want ChatGPT or Claude to assist me, granular details about the plan and intended results, plus anything else which may be useful. (Such as using UK English!)
- Importantly, you can upload files (or in the case of Claude link directly to those in your Google Drive) for constant reference on your project work. This allows you to provide existing documents such as a client’s strategy plan, brand positioning, list of key stakeholders – you name it.
- Conversations, drafts, and attachments all live in one place: no more frantic searching or accidental “where did I say that?” moments. ChatGPT and Claude can then access and refer to this data, creating a handy one-stop knowledge bank.
- When it’s time to revisit an ongoing piece of work, I pick up exactly where I left off, whether I’m in ChatGPT or Claude. For good hygiene, I’ll rename the conversations so I can easily spot where relevant work exists.
- I also can then create certain conversations to quickly help: such as a ‘jargon buster’ where I have set instructions so that I only need to copy and paste an unwieldy acroynym or niche detail, and it will comb through the archive to detail its meaning and specific relevance.
ChatGPT folder creation: Getting started takes seconds
---
## How I Use Folders Across ChatGPT and Claude
### **1\. Folder Structures**
I use the same structure and naming conventions in both platforms. For example:
- “Faur – Applied Comms AI”
- “Client Name – Press Briefings”
- “Newsletter Drafts – 2025”
This makes switching between tools seamless. If I start a strategy outline in ChatGPT but want to use Claude for further ideation, everything is easy to locate and reference.
For larger client projects, I will create several folders for distinct workstreams, whereas for potential new clients or smaller pieces of work, I can create one folder as a singular point of focus.
Claude folder creation: also a simple setup
### **2\. Segmentation by Workstream**
I separate folders by:
- Client/project (consulting, retainer work, etc.)
- Experiments and R&D (prompt building, tool reviews)
- Content creation (articles, newsletters, social posting)
- Learning and reference (AI guides, research digests)
At the same time, I do often like to keep folders generally relevant, rather than getting too niche – it can be helpful for cross-referencing and ensuring consistency.
### **3\. Collaborative Handover**
When sharing outputs or briefing notes, it’s far easier to locate and copy everything when it’s grouped together. Even for solo work, knowing where everything lives reduces friction and mental load.
---
### Practical Tips for Power Users
- **Use Consistent Naming:** Start folder titles with the client or project name, then the type of work. E.g., “Client X – Stakeholder Mapping.”
- **Regular Reviews:** Archive or delete folders you’re no longer using—AI workspace clutter can build up fast. Also ensure that the documents and project instructions continue to be relevant, deleting or replacing as necessary.
- **Backup Important Threads:** Since folders don’t sync across platforms (yet), export or copy vital information between ChatGPT and Claude as needed.
- **Make Folders Your Default:** Even for “just testing” ideas, drop them into a dedicated folder. You never know when you’ll want to revisit an experiment, plus you can then prompt to surface random and/or relevant ideas from the past.
---
## What’s Still Missing?
Project Folders are a big step forward, but there’s more to do:
- **Folder Organisation:** Currently it’s just a big old list, rather than being able to have folders within folders, and introduce more structure.
- **Permissions and Sharing:** Right now, folders are personal. It would be brilliant to see shared folders for collaborative projects in future updates.
- **Attachments & Rich Media:** Handling files is improving, but neither platform has nailed seamless file organisation within folders yet. Popular formats such as Word and PDFs are okay, but it can become a bit of a lottery beyond those.
- **Cross-Platform Sync:** Currently, there’s no way to sync folders between ChatGPT and Claude. If you use both, you’ll need to manually mirror your organisation.
---
## Why This Matters (Especially for Comms Pros)
Communications, PR, and consulting work is inherently multi-threaded and fast-moving. You’re juggling clients, campaigns, brainstorms, and a ton of raw and partially formed ideas. If your AI workspace is a mess, you’ll waste time and lose momentum.
By using Project Folders on ChatGPT and Claude – or though any other Gen AI option which provides the functionality – you create order, speed, and focus. You’ll find it easier to deliver polished work, revisit old ideas, and demonstrate value to clients (or your own team).
---
## Final Word: A Simple Step Forward, With Substantial Benefits
AI tools continue to become more powerful, but sometimes, it’s the simple (and classic!) features that deliver the most value. Folders sound basic, but for anyone working with AI at scale, they can be transformational. They help you move beyond scattered experiments and into sustained, professional, efficient and effective use.
If you haven’t set up folders on both ChatGPT and Claude yet, I recommend carving out 10 minutes to do it now. There is plenty of room for improvement, but it provides a substantial step closer to these tools becoming the helpful AI assistant who is consistently on the ball, and sometimes ahead of the game. You’ll thank yourself the next time you need to quickly draft up a great new idea for an impending client meeting, or pick up a project right where you left off.
---
**Found this useful?** Subscribe – if you haven’t already – and please forward this to a colleague who may be facing a similar challenge.
*Applied Comms AI is powered by* [*Faur*](https://faur.site/)*. We help organisations navigate the intersection of communications and technology.*
---
### Building a LinkedIn content creator with Claude, giving Grammarly a reality check, and a (welcome) audit by ChatGPT’s Deep Research
URL: https://www.appliedcomms.ai/newsletter-2025-07-08/
Last updated: 2026-02-25T16:17:19.000Z
*The warmest of welcomes to the *Applied Comms AI newsletter*. There’s plenty to get stuck into, and I hope you’ll find something of use below – this is the first ‘fully featured’ edition, so any and all feedback welcome.*
*In this edition, we dive headfirst into Claude's code-free Artefact app builder to see if we can build a proper LinkedIn content creation tool. Plus, a reality check on whether Grammarly's AI features actually help comms pros, and how ChatGPT’s Deep Research feature may save you hours of time – even if it doesn’t measure up to industry-specific technical expertise. Well, not yet, anyway…*
*– Michael MacLennan (*[*connect on LinkedIn*](https://www.linkedin.com/in/maclennanmichael/)*)*
# 1\. App Building: LinkedIn Content Creator
*In each edition, we build something new with AI to test what's actually possible (versus what the hype suggests).*
**What We Built:** A LinkedIn content ‘wizard’ that turns basic ideas into properly formatted posts, with a magical theme for added user-friendliness
**App Purpose:** Imagine building an app for your client/boss which easily generates appropriate posts for them, saving hours of time
**Technical Purpose:** Testing Claude's newly capable Artifact feature for building interactive apps without traditional coding
**What happened:**
- It was fairly easy to build a functional tool in under 3 hours that generates LinkedIn posts with ideas generation, audience focusing, and tone-of-voice options
- The interface is surprisingly slick: including proper forms, real-time preview, the ability to refine results
- However, there are some notable limitations around advanced customisation and data persistence
**The Reality Check:** Claude’s Artifacts has great potential for rapid prototyping and basic functionality, but you’ll still require extra technical expertise to go beyond this. You’ll also be frustrated at the ability to directly edit code, if you do have that level of knowledge
**Try It Out:** Have a play with [the Applied Comms AI: LinkedIn Post Wizard](https://claude.ai/public/artifacts/f7cbf323-5739-42cb-abff-76b847ce9515), and let me know how you would improve upon it
[READ THE FULL BREAKDOWN & TAKEAWAYS: ‘How good is Claude at building an AI Tool for LinkedIn Content Creation?’](https://www.appliedcomms.ai/app-building/claude-artifact-app-builder-linkedin-content-creator/)
*We're* [*conducting our App Building in public*](https://www.appliedcomms.ai/app-building/) *so you can learn from both our wins and mistakes.*
---
# 2\. Tool Review: Grammarly
*Our tests of AI-powered apps and software – what’s worthy of inclusion in your comms toolkit?*
**What we’re testing:** Grammarly, the leading AI-enhanced writing assistant
**The Promise:** AI that catches tone issues, suggests better phrasing, and adapts to your brand voice
**In Practice:** If you're a comms pro writing across email, social, web, and documents, this is your insurance policy, and the closest you’ll come to a personal subeditor. At Faur, where I've been juggling multiple client communications daily while building a new business, it's been indispensable. However, as with other generative AI tools, it does attempt to flatten the language you use, with a bland tone which won’t help you stand out from the crowd
**Best for:** Ensuring you are unlikely ever again to write an important message – or publish a headline – with a mortifying typo
**Skip if:** You're looking for sophisticated AI writing assistance, or have experienced copywriters who know exactly how to nail voice
**Verdict:** Reliable for basics, overhyped for advanced features
[CHECK OUT THE FULL REVIEW: ‘Grammarly Review: Your Everywhere Comms AI Subeditor’](https://www.appliedcomms.ai/tool-testing/grammarly-review/)
---
# 3\. Prompt Engineering: Deep Research for Organisational Briefs
*Your copy-paste ready AI prompt, tested and refined through multiple iterations.*
**Use case:** Making the most of [ChatGPT’s independent Deep Research agent](https://openai.com/index/introducing-deep-research/) for in-depth communications planning
**Where to go:** [Visit ChatGPT,](https://chatgpt.com/) and click on ‘Run deep research’ from the Tools options at the bottom of the prompt box
**The Prompt:**
```
You are a strategic communications researcher. I need you to analyse the following information and provide a comprehensive brief.
Context: [Insert your topic/issue]
Sources: [Paste your research materials]
Please structure your analysis as:
1. SITUATION SUMMARY (3-4 sentences)
2. KEY STAKEHOLDERS (who matters and why)
3. MAIN NARRATIVES (what stories are being told)
4. RISKS & OPPORTUNITIES (what could go wrong/right)
5. STRATEGIC RECOMMENDATIONS (3 specific actions)
Focus on practical insights a communications team can act on immediately. Avoid generic advice.
```
**Pro tip:** Once you’ve entered your prompt, ChatGPT will ask some follow-up questions prior to commencing the research – take the time to consider and answer these as comprehensively as possible. It’ll only take a couple of extra minutes of your time, and will help tailor the end results to your needs
**Example:** To properly test this tool for Applied Comms AI, I decided to turn the lens on my own organisation, [Faur](https://faur.site/?ref=appliedcomms.ai). Through a tailored version of the prompt above, after an hour I received a helpful 10-page, 5900-word audit covering everything from competitive positioning to SEO analysis, with actionable recommendations
- **FIND OUT MORE :** [**'We Tested ChatGPT's Deep Research Tool: Here's What Comms Professionals Need to Know’**](https://www.notion.so/Newsletter-2-21f94b5ae08880dca666ca2e1cf98bb1?pvs=21)
*We test every prompt we share – this one's been refined through multiple iterations*
---
# Comms AI in the Headlines
News of note this week:
- **Cursor pricing backlash**: The leading developer-focused AI tool has faced a user exodus over subscription changes – a handy reminder that AI tool adoption isn't guaranteed, even when the tech works ([link](https://www.notion.so/Newsletter-2-21f94b5ae08880dca666ca2e1cf98bb1?pvs=21))
- **Gallup AI in comms report**: New research shows that 73% of comms professionals are using AI, but only 31% have formal policies – worth reading for benchmarking your team's approach ([link](https://www.notion.so/Newsletter-2-21f94b5ae08880dca666ca2e1cf98bb1?pvs=21))
- **Agencies create AI search units**: Comms and marketing agencies are building specialist teams for AI-powered search optimisation – is it time to shift some of your SEO budget? ([link](https://digiday.com/marketing/agencies-create-specialist-units-to-help-marketers-solve-for-ai-search-gatekeepers/))
- **Xbox producer tells staff to use AI to ease job loss pain**: A lesson in how not to speak about AI to an internal audience ([link](https://www.bbc.co.uk/news/articles/ckglzxy389zo))
---
# Closing & CTAs
Thanks for reading – and also for all the enthusiasm and kind words since I launched Applied Comms AI. We’re still at an early stage (I already consider us a community!), and so I cherish your thoughts and opinions even more than usual.
The hope is that each newsletter and article provides practical support and inspiration, as much for what *not* to do, as to how to proceed. Let me know if the balance feels right.
**Your turn:** How are you using AI in your comms work? Hit reply – I read everything, and we’ll be shaping future editions according to your insights and needs.
**Worth sharing?** Forward this to a colleague who's trying to make sense of AI.
**New here?** Subscribe for weekly insights 👇🏻
## Sign up for Applied Comms AI
Practical AI insights for forward-thinking comms leaders.
Subscribe
Email sent! Check your inbox to complete your signup.
No spam. Unsubscribe anytime.
### How good is Claude at building an AI Tool for LinkedIn Content Creation?
URL: https://www.appliedcomms.ai/claude-artifact-app-builder-linkedin-content-creator/
Last updated: 2026-02-25T16:19:05.000Z
*I used Claude’s latest app-building capabilities to create a tool that aims to offer several leading LinkedIn posting approaches as part of a structured, AI-powered workflow designed to save time and enhance engagement. Let’s delve into whether it was all worth it…*
## Applied Comms AI’s App Building: An Intro
…but before we get ahead of ourselves, a quick bit about the purpose. At Applied Comms AI, I'm documenting the messy reality of integrating AI into communications work: not just theorising about it. That means building actual tools, testing them properly, and sharing what works (and what absolutely doesn't).
Apps in particular is a fascinating area for me: there are now a variety of providers who promise that, through AI, people with limited to no knowledge of coding can build capable tools and apps. So, is this true, and if so what are the purposes and implications for AI? That’s what we intend to find out in this section, and within each article I’ll look to experiment with a diverse variety of providers, approaches, and challenges.
We will explore the options on offer, starting simply. In great timing for this kick-off, [Anthropic just announced the ability](https://www.anthropic.com/news/claude-powered-artifacts) ‘ to build, host, and share interactive AI-powered apps directly in the Claude app’, and then to be able to share them.
This feels like a particularly straightforward and easy-to-implement option for app building, so how effective is it in practice?
# **🎯** The Problem We're Trying to Solve
**In brief:** Anyone who posts regularly on social media will have that moment of staring at a blank prompt window and being stuck at how to get started. That’s just as true for comms professionals seeking to create great content that engages the target audiences for their organisation/client.
**What we're building:** A magical, wizard-style interface that guides users through creating LinkedIn posts by systematically selecting post types, tones, audiences, and content approaches, then uses AI to generate multiple variations optimised for engagement.
**Why it matters:** Let’s say 2-4 hours is spent weekly on LinkedIn content creation. A structured approach could reduce this to 30 minutes by producing strong first drafts that tick all the essential boxes mentioned above – and which can then increase post quality and engagement rates.
**Reality check:** This won't replace strategic thinking or genuine insights: it's a framework tool that handles the mechanical aspects of LinkedIn optimisation, leaving humans to focus on strategic approaches, authentic messaging, and relationship building.
# **🧪** The (Testable) End Result
This is slightly spoiler-ific, but I prefer to present the result(s) up top, so you don’t have to wade through the rest of the detail unless you want to. So without further do:
- [Try out the Applied Comms AI: LinkedIn Post Wizard](https://claude.ai/public/artifacts/f7cbf323-5739-42cb-abff-76b847ce9515)
Take a look, see what you think, and please do let me know. I detail the process I went through below, though first…
# ⚠️ Reality Check: Before You Start Building
- **Time investment**: 3-5 hours minimum (including debugging and iterations)
- **Technical skills needed**: Basic understanding of prompting AI, patience with troubleshooting
- **Success rate**: Expect 2-3 failed attempts before getting something functional
- **Best case scenario**: You'll create a useful tool that saves 30 minutes (or more) per week
- **Worst case scenario**: You'll spend an afternoon learning why most people buy rather than build
- **Ask yourself first**: Could I achieve 80% of this benefit by simply creating a saved prompt in ChatGPT? If yes, start there.
# 🤔 The Initial Plan (And Why We Might Be Wrong)
**Our hypothesis:** By systematising the proven elements of high-performing LinkedIn content (post types, tones, audience targeting), we can help communications professionals create more engaging content in significantly less time.
**Success looks like:**
- Users generate usable LinkedIn posts within minutes, saving considerable time umm-ing and ahh-ing
- There’s a level of customisation at play, so the user is still playing an active role
- Timely and relevant posts can be generated, with the user able to pin-point what they’d like to talk about
**And catastrophe looks like:** If the app fails to work, feels like a novelty, and/or isn’t intuitive and fun to use.
**Technical approach:** To ensure it didn’t feel too dry for potential clients or teammates to use, I opted for a whimsical "wizard" interface that makes professional content creation feel engaging rather than clinical.
- **Platform:** React with Claude AI integration
- **AI model:** Claude Sonnet 4
- **Key features:** Quick Mode for rapid generation, Step-by-Step wizard for guidance, A/B testing with multiple variations
# 🧱 The Production Process
**What we did:** Claude’s apps are labelled Artifacts, [with a specific section](https://claude.ai/artifacts) that allows you to create your own artifacts, providing inspiration for these. It then prompts you to pick a category or build an idea from scratch.
I went to ChatGPT to provide inspiration for five different tones of voice for LinkedIn posts, also considering different types of audiences.
A neat thing about the creation process was that Claude outlines the plan in advance, which provided another opportunity to refine and improve upon this.
Once the plan was agreed, it got to work in the background, allowing you to view code as it is written/rewritten.
This is where I ran into the first hitch, which is that the app… didn’t work, sticking on a problem which Claude was unable to resolve. This [is very common to the 2025 phenomenon of vibe coding](https://medium.com/data-science-in-your-pocket/dont-be-a-vibe-coder-30fa7c525971), where people without coding knowledge find themselves stuck at a certain stage of the project, and unable to resolve themselves.
**My options**:
1. Start completely over (what I did)
2. Copy the code elsewhere and manually debug (requires actual coding skills)
3. Give up and accept that AI app-building isn't ready yet
**The uncomfortable truth**: "Vibe coding" – where non-programmers use AI to build applications – works brilliantly until it doesn't. When you hit that wall, you're stuck in no-man's land: too complex for simple fixes, too basic for serious development tools.
**What this means for comms professionals**: Unless you're comfortable with the possibility of losing hours to technical dead-ends, approach AI app-building as an experiment, not a business solution. The tools are impressive but unpredictable.
This is particularly an issue with Claude’s artifacts, since they don’t currently let you amend the code yourself. In this case I resolved by first going through the app building stage again from scratch, simplifying some of the features that it would have, and then running into a similar issue again and prompting it several times to resolve.
Eventually it did, and though this was frustrating, in total it added no more than 30 minutes or so – not significant in the grand scheme of things.
# 🔧 The Functionality (aka Does It Actually Work?)
After the extra efforts mentioned above, we had a simple and functional app which was indeed generating posts.
Able to generate potential topics at the touch of a button, in this case I prompted it into a topical hot take aimed at corporate leaders, about [the recent pricing backlash faced by Cursor](https://analyticsindiamag.com/ai-features/cursors-pricing-backlash-sparks-developer-exodus/).
Once generated, the app allows you to refine the post. In this example, I looked to both make it shorter and add a more casual tone.
## **⚖️** The Verdict – and Potential Uses
- **Overall verdict:** As a first app-building attempt for Applied Comms AI, this felt like an easy gateway into potentially building more complex comms apps (or ‘Artifacts’ as Anthropic will no doubt insist you call them). It’s already a solid time-saving app, with plenty of potential for iterating upon
- **Time spent:** Approximately 3 hours (including initial prompt generation, going through a first failed attempt, refining and iterating upon the final version)
- **Who it's best for**: In-house comms teams managing executive thought leadership programmes or agencies running multiple client LinkedIn accounts: it systematises content creation while maintaining brand voice consistency across stakeholders
- **What it does well**: It provides a straightforward user interface for handling the mechanical aspects (format, engagement hooks, A/B variations)
- **What I particularly like:** The playful wizard interface was easy to add as a refinement, and feels as though it would reduce the intimidation factor for clients/other colleagues, aiding uptake and usability
- **What I particularly *dislike*:** At this time, it can't replace genuine insights or an authentic voice. It’s more of a sophisticated content assistant, rather than a creative replacement
- **What's still missing**: Automated URL fetching for news commentary (not available in Claude’s demo environment), LinkedIn API integration for direct publishing to become truly workflow-ready (also not available), the ability to upload helpful documents (such as for tone of voice, positioning, etc)
- **How *you* can build on this**:
# **💭** The Wrap-Up
Currently, Claude’s Artifacts app-building capability provides potential, but lacks some of the features that would be needed for a more customised and client-ready product.
The LinkedIn Post Wizard built for this article is effective enough as a sophisticated content assistant that handles structure and optimisation, freeing communications professionals to focus on strategy and authentic messaging.
The wizard won't make a boring company announcement fascinating, but it will help you present it in the most engaging way possible for a generic audience type, with it easy enough to tailor a similar Artifact unique to your purposes – if you want to use the steps above as inspiration.
# ❓What I'd Love to Know
**Try the tool** and let me know: Does it actually save you time, or does crafting the inputs take longer than writing the post yourself?
**Share your approach**: Are you building AI tools internally, buying existing solutions, or avoiding the whole thing? Hit reply and tell me what's working (or what isn't).
**Vote for future experiments**: What should I build and test next?
- Crisis communications response templates
- Internal comms tone analyzer
- Media pitch personalization tool
- Something completely different (suggest it)
The goal isn't to become an AI development agency – it's to understand which applications genuinely help communications work and which are expensive distractions.
---
**Found this useful?** Subscribe to [the Applied Comms AI newsletter](https://www.appliedcomms.ai/) – if you haven’t already – and please forward this to a colleague who may be facing a similar challenge.
*Applied Comms AI is powered by* [*Faur*](https://faur.site/)*. We help organisations navigate the intersection of communications and technology.*
---
### Grammarly Review: Your Everywhere Comms AI Subeditor
URL: https://www.appliedcomms.ai/grammarly-review/
Last updated: 2025-08-08T13:28:43.000Z
## Should You Care?
- **Rating:** 4/5 stars
- **Best for:** Comms professionals – and pretty much any digitally communicative worker – multiple platforms daily
- **Monthly cost:** Free tier available; Pro plans from £12/month ([info here](https://www.grammarly.com/plans))
- **Time to value:** Minutes (install its extensions for the software/apps you use, and it immediately starts working as you type)
- **For The Skimmers:** After 18 months of daily use across almost every conceivable platform, Grammarly has become my AI-powered digital safety net for spelling mistakes and grammatical howlers. It's brilliant for catching embarrassing errors before they reach clients, and the Chrome/Safari/Desktop extensions mean it's always there. Though beware: it can flatten your writing style if you accept every suggestion. For busy comms teams, it's insurance against reputation-damaging typos.
Reviewing Grammarly as it's reviewing me: oh, the irony!
## The Test Parameters
**What we tested:**
- Plan: Having used the Free plan for most of this year, I went back on the Premium plan in the past month (I missed its sentence-rewriting capabilities to help when I have to fire something in a mad dash)
- Use cases examined: Client emails, newsletter writing, social media posts, formal proposals, website copy (you name it, really)
- Comparison baseline: Raw writing speed and error rates without assistance (I’m a touch typist of the most haphazard order)
**What we didn't test:**
- Business/Enterprise team features (solo user experience only)
- Mobile app functionality (desktop-focused testing; I have used the iPad app a bit, but not enough to be able to critique)
- Catching accidental plagiarism (since using it for my own writing)
- Generating text with AI prompts (I go elsewhere for drafting and content creation)
## First Impressions
### Setup Experience
- Time from signup to first output: Under five minutes
- Onboarding process: Install the extension(s), log in, done
- Initial confusion points: None, refreshingly
- "Aha" moment: Likely the first time it caught a misused word (probably something along the lines of "manger" instead of "manager" in a client email) or improved sentence construction, and I realised it was far more helpful than the typical Word document spell/grammar checker
### The Interface
- First reaction: "Oh, it's just... there. Everywhere."
- Navigation: What navigation? It's ambient—red underlines appear, you hover, you fix
- Mobile experience: Not tested extensively
- Comparison to similar tools: Unlike Word's spell check, this follows you everywhere
The beauty is its invisibility. There's no dashboard to navigate or app to open. The Chrome extension just quietly underlines your mistakes with those familiar red squiggles. Hover over them, and you get suggestions. It's the digital equivalent of having a proofreader looking over your shoulder—but less creepy.
## Test Case
Here's how it performs in practice, taking a meta approach to how it was used in this very review.
Tailoring your Tone: The Grammarly app/website allows you to set goals
### Test Case 1: Writing This Very Review
**The brief:** Create a 1,000+ word tool review that's informative but doesn't bore its readers to tears. (Hey, you’ve already made it most of the way through…)
**The process:**
1. Started drafting in Notion (Grammarly following along via [the Mac desktop extension](https://www.grammarly.com/desktop/mac))
2. Caught "took" when I meant "tool" – oh, will I ever learn?
3. Other corrections suggested throughout, such as flagging a rambling sentences that even I couldn't follow
4. Then imported the first draft into the Grammarly app, where it immediately prompted me to set goals (see screenshot above)
5. After indicating my goals, it provided review suggestions in a right-hand window. \[Insert thoughts on these once done\]
**The output:** This review: far cleaner than it would have been otherwise, but still recognisably mine (as mentioned, I did reject some suggestions for this reason)
**The verdict:** As a former newspaper subeditor myself, it’s the closest I have had – and will ever get – to have a personal subeditor, providing a hugely helpful safety net for solo working and sending important emails/documents to clients.
**Time saved/lost:** In this case, probably at least 10-15 minutes saved on a an extra draft revision or two, plus the smug meta satisfaction of testing a tool while writing about testing it
Review suggestions: Grammarly checks for Correctness, Clarity, Engagement, and Delivery
## The Good, The Bad, The Ugly
### What It Nails
- **Ubiquitous presence:** Works in Gmail, LinkedIn, Notion, Ghost… pretty much everywhere I write. It really does feel like an ever-present assistant.
- **Speed:** Instant feedback means fixing errors in real-time, not in retrospect
- **Context awareness:** Now detects more complex errors, including mistakes in advanced sentence structures
- **Tone detection:** This can be helpful when switching between formal proposals and casual team chats, although the results can feel a bit broad – even when this is the case, it can still be helpful to consider and personally revise
### Where It Falls Short
- **Style homogenisation:** Much like any generative AI tool, accept too many suggestions and your writing will become vanilla. Much like ChatGPT or Claude, there are certain words and phrases that it appears moderately obsessed by…
- **Overcorrection:** …And on that note, the more personality you insert, the more it can flag intentional stylistic choices as errors
- **Extension quirks:** Sometimes obscures buttons or form fields on websites, which can provoke mini outbursts of frustration. However, never enough to have removed said extensions!
- **Limited customisation:** The extension lacks the tone controls available in the web version
### Deal Breakers (If Any)
- There are other highly rated Grammarly alternatives out there ([see this article for some of the contenders](https://kindlepreneur.com/grammarly-alternatives/)), so worth doing your research
- Perhaps most notably, Apple seemed to launch a [‘Grammarly killer’](https://www.cnet.com/tech/services-and-software/did-apple-intelligences-rewrite-tool-just-kill-grammarly/) last year through its own AI-powered writing tool. However, despite the hype, so far this is more limited in scope and fairly buried. Of course, it’s likely to improve and become more accessible over time, but I wouldn’t choose it over Grammarly at the moment.
- The premium price might sting for solo freelancers (though the free tier is generous)
- If you're precious about your unique writing voice, the constant suggestions might grate
## Who Should Use This
### Perfect For:
- In-house comms teams writing across multiple platforms daily
- Agency professionals crafting client communications at speed
- Anyone whose reputation depends on error-free writing
### Skip If:
- You don’t need to worry too much in your daily role about mistakes which won’t be caught by the standard spell-checker
- Your team has dedicated proofreaders for everything
- Budget is tight, and the free tier's more limited features aren't enough
### Sweet Spot Scenario
If you're a communications professional writing 20+ pieces weekly across email, social, web, and documents, and you've ever sent an important message with a mortifying typo, this is your insurance policy. At Faur, where I've been juggling multiple client communications daily while building a new business, it's been indispensable.
## Quick Tips
1. **Disable on specific sites:** Turn it off on your CMS if it interferes with editing
2. **Personal dictionary:** Add client names and industry jargon to avoid false flags
3. **Selective acceptance:** Read suggestions critically—don't auto-accept everything
## The Bottom Line
**In a single sentence:** Grammarly isn't revolutionary: it's just reliably there when you need it, as the conscientious subeditor you never knew you needed.
## Tool Information Box
- **Tool:** Grammarly
- **Website:** [grammarly.com](https://www.grammarly.com/plans)
- **Pricing:** Free tier; Premium currently £10/month (billed annually)
- **Free trial:** Yes (through the free tier, no time limit)
- **Category:** Writing Enhancement/Proofreading
- **Review last updated:** 7 July 2025
*Using Grammarly already? What's been your experience—salvation or style-killer? Let us know your thoughts in the comments below.*
### We Tested ChatGPT's Deep Research Tool: Here's What Comms Professionals Need to Know
URL: https://www.appliedcomms.ai/deep-research-comms-test/
Last updated: 2026-04-14T07:35:40.000Z
ChatGPT's [Deep Research tool launched in February](https://openai.com/index/introducing-deep-research/) with a bold promise: AI that can "basically do all your research for you." For communications professionals drowning in competitive intelligence, crisis monitoring, and endless stakeholder mapping, this sounds like either salvation or snake oil.
We decided to find out which.
## What Deep Research Actually Does
Think of Deep Research as hiring an inexhaustible research assistant who works for several hours straight, systematically browsing hundreds of sources to create comprehensive, cited reports. *Unlike* ChatGPT's regular search function, which gives you quick answers in seconds, Deep Research deliberately takes time to deliver thorough, documented analysis.
The tool uses OpenAI's o3 reasoning model to conduct multi-step research projects, analysing and synthesising information from diverse online sources. Every output comes fully documented with clear citations – crucial for maintaining professional credibility in comms work.
The results themselves adhere closely to the type of research and analysis reports that agencies spend days or weeks preparing for clients, charging accordingly. So how does it compare?
**Current availability (**[**full pricing here**](https://openai.com/chatgpt/pricing/)**):**
- **Free users**: 5 lightweight queries per month (“Lightweight” = more cost-efficient model based on o4‑mini)
- **Plus subscribers (£20/month in the UK)**: 10 full searches + 125 lightweight queries per month
- **Pro subscribers (£200/month in the UK)**: 125 full Deep Research queries per month (plus also 125 lightweight queries per month)
## The Comms Professional's Reality Check
What does this actually mean for day-to-day work?
**Where it could genuinely help:**
- Competitive intelligence gathering
- Crisis background research
- Industry trend analysis
- Stakeholder sentiment mapping
- Client briefing preparation
**Where you'll still need human judgement:**
- Verifying claims and sources (you’ll need to review first before sharing!)
- Understanding context and nuance
- Strategic interpretation
- Relationship-based insights
To its credit, OpenAI is transparent about limitations: Deep Research can sometimes hallucinate facts, struggle to distinguish authoritative information from rumours, and often fails to convey uncertainty accurately.
## Our Test: Deep Researching Ourselves
To properly test this tool for Applied Comms AI, we decided to turn the lens on ourselves. We asked Deep Research to conduct a comprehensive audit of our own organisation, [Faur](https://faur.site/): analysing public perception, competitive positioning, digital presence, and identifying opportunities and gaps.
**Here's the prompt we used.** We undertook several revisions to ensure that it is helpful in providing both a purpose and overall structure useful for organisational analysis/peer review **(please feel free to copy, paste, and amend for your own use)**:
```
"I'm testing Deep Research for a newsletter about AI in communications.
Please conduct a comprehensive analysis of 'Faur' - a UK communications
consultancy founded in 2024 by Michael MacLennan. Research:
PUBLIC PERCEPTION & POSITIONING:
- How Faur is described across websites, press coverage, and networks
- The founder's professional reputation and thought leadership
- Stated expertise areas and service offerings
- Client work mentioned publicly
COMPETITIVE LANDSCAPE:
- Positioning vs other UK communications consultancies
- Unique selling propositions compared to traditional agencies
- The 'collective' model vs conventional structures
DIGITAL PRESENCE:
- Website effectiveness and messaging clarity
- Social media presence and engagement
- Content strategy and thought leadership output
OPPORTUNITIES & GAPS:
- Areas where competitors are stronger
- Underutilised positioning opportunities
- Potential reputation risks or weaknesses
Please provide specific examples, URLs, and dates where possible."
```
## The Process: What Actually Happens
When you select "Run Deep Research" from the Tools menu, ChatGPT typically asks clarifying questions before starting. In our case, it asked three smart follow-up questions:
1. **Competitor comparisons**: Did we want named competitors analysed, or general market positioning?
2. **Geographic scope**: UK-focused or international sources?
3. **Digital presence priorities**: Specific platforms or general overview?
This clarification step is a helpful prompt in itself – it forces you to think through research boundaries and potential sensitivities. We opted for general positioning (since this was going public), UK focus, and broad digital analysis.
Once it begins, real-time progress is shown through the Activity panel on the right. You can watch in real-time as it:
- Identifies research angles and sub-questions
- Searches across different types of sources
- Builds up a comprehensive source list
- Synthesises findings into structured insights
The AI confirmed our parameters: *"I'll focus on how Faur and its founder, Michael MacLennan, are discussed online, their market stance, and visible strengths or gaps"* before diving into the research.
**Time investment:** Our first attempt unexpectedly got stuck on a hard-to-access source (see below). After giving it some time, we stopped and undertook a second attempt – this time with clearer boundaries around time limits, since even AI research assistants can clearly get lost down rabbit holes… If you account for refining the prompt, then going away while it was doing its thing, the process probably took 10-15 minutes of time.
**Source quality:** The tool pulled from a mix of professional networks, industry publications, company websites, and news sources. Impressively, it found several mentions we weren't aware of, although it also missed some obvious sources.
## What We Actually Discovered
**The unexpected challenge:** Our first research attempt got stuck trying to access a Facebook post about a speaking engagement. After the best part of a day of the AI trying to access restricted content, we had to restart with clearer instructions: "skip anything after a reasonable time period if too difficult to access."
This reveals something important: Deep Research can get overly persistent with inaccessible sources, burning through your allocated time. Always include time boundaries in your prompts, so it can be ready for you closer to when you’re after it.
**The surprisingly comprehensive:** Within the space of 61 minutes (it allows you to go away and let’s you know when finished), Deep Research had conducted 53 searches, referred to 16 sources, and produced a 10-page, 5900-word audit covering everything from competitive positioning to SEO analysis.
Delving into details including client testimonials, speaking engagements, and my own ScotlandIS board appointment, the structured approach (Public Perception, Competitive Landscape, Digital Presence, Industry Standing, Opportunities & Gaps) was methodical and thorough.
**The useful insights:**
- **Positioning clarity**: Identified that our "collective model" differentiator could be explained more clearly to potential clients
- **Content gaps**: Spotted that we have only one detailed case study publicly available despite working with multiple high-profile clients
- **SEO opportunities**: Noted our unique name "Faur" helps with branded searches, but we're missing generic search traffic
- **Competitive angles**: Highlighted how our AI-focused content positions us differently from traditional agencies
**The concerning accuracy issues:**
- Some details were outdated or slightly wrong (mixing up dates, over-stating follower counts)
- Occasional misattribution of achievements or quotes
- Limited understanding of UK market context and Scottish business landscape – the level of knowledge in terms of the competitive analysis felt a bit too surface-level compared to what a sector expert would lay out
- As a similar note, missed nuanced relationship knowledge that comes from actually being in the industry
**The strategic recommendations:**
- Suggested emphasising "Scotland base with global reach" as a differentiator from London agencies
- Recommended more explicit positioning as an "AI-informed communications consultancy"
- Identified potential reputation risk of being too founder-dependent
- Proposed specific content strategies like featuring client logos (with permission) on the homepage
- *These recommendations felt valid, if already fairly evident – there’s nothing there that I hadn’t personally thought of prior to now (with various dull reasons – generally relating to lack of time – as to why I haven’t acted on them)*
## Practical Prompting Tips for Using ‘Deep Research’ for Comms
Based on our testing, here's what works:
**Be specific about scope:** "Analyse reputation over the last 12 months" works better than "research our reputation". If feeling unclear on what you’re after, you can even ask ChatGPT to help you shape a more specific and useful scope, prior to conducing the Deep Research
**Structure your request:** Break complex research into clear categories (perception, competition, opportunities)
**Request examples:** Always ask for "specific examples, URLs, and dates where possible"
**Set context:** Explain why you're researching and how you'll use the findings
**Think through sensitivities:** Consider what you're comfortable making public if this research gets shared
**Set time boundaries:** Include phrases like "skip sources that take longer than reasonable time to access" to prevent getting stuck
**Expect clarifying questions:** The AI will likely ask 2-3 follow-ups to refine scope – this is helpful, not annoying
**Plan verification:** Build in time to cross-check key claims and sources – the initial results should certainly not be considered client- or stakeholder-ready. Human oversight is still all important
## When to Use Deep Research (And When Not To)
**Perfect for:**
- Quickly assembled competitive audits to kick off project work
- Client sector research and briefing prep (picking up details that go beyond typical searching)
- Crisis background research requiring multiple sources
- Industry trend analysis across publications
- Stakeholder sentiment mapping
**Skip it for:**
- Quick fact-checking (regular search is faster)
- Relationship-based intelligence (it can't read the room)
- Real-time crisis monitoring (it can take hours to deliver, and relies more on historical information)
- Highly technical or niche industry research
- Information requiring insider knowledge
## The Verdict: Impressively Thorough, Strategically Valuable, Though Not as ‘Deep’ as Claimed – Yet
Deep Research delivered something truly useful: a comprehensive external perspective on our own business that would have taken days to compile manually, and that we simply wouldn’t have the resource to manually produce. The 10-page audit covered positioning, competitive landscape, digital presence, and strategic opportunities with a level of systematic thoroughness we could never have achieved in the same timeframe.
**What impressed us:** The AI identified our unique positioning elements (the collective model, Scottish base with global reach, AI-forward approach) and articulated them clearly. It spotted content gaps, SEO opportunities, and competitive differentiators, and packaged them up neatly before providing actionable next steps.
**Where human oversight proved crucial:** While comprehensive, the research contained some minor factual errors, outdated information, and missed contextual nuances that come from actually working in the UK communications market. Some insights were spot-on; others would need correction before presenting to others.
**The strategic value:** Despite accuracy issues, the external perspective revealed blind spots and articulated opportunities we hadn't considered. The recommendation to position more explicitly as an "AI-informed communications consultancy" was particularly helpful – it's something we'd been doing but not saying clearly enough.
For Applied Comms AI purposes, this demonstrates Deep Research's sweet spot: comprehensive initial research and strategic thinking acceleration, with essential human verification and context-setting.
## Your Turn: Deep Research Prompts to Try
If you're ready to experiment, here are three comms-focused prompts worth testing:
**Crisis Preparedness Audit:**
```
"Research potential reputation risks for [Your Organisation/Client] including:
recent industry controversies, common criticism themes, regulatory issues,
and emerging challenges that could impact reputation in the next 12 months."
```
**Competitive Intelligence Deep Dive:**
```
"Analyse [Organisation]'s competitive positioning vs [3-5 competitors].
Focus on: messaging strategies, share of voice, thought leadership content,
and opportunities for differentiation."
```
**Stakeholder Sentiment Mapping:**
```
"Research stakeholder perceptions of [Organisation] across: employee
sentiment, customer feedback, media relationships, and industry peer
recognition. Identify recurring themes and improvement areas."
```
## The Bottom Line
Deep Research isn’t ready to replace strategic communications expertise – yet – but it can significantly accelerate the research phase of our work. At £20/month for 10 full searches, it's massive level of value given the many hours of extra time it would take even for a single search.
Just remember: trust, but verify. And always apply the strategic context that comes from actually understanding the communications landscape.
*Have you tested Deep Research for comms work? Hit reply and share your experiments – we might feature insights in a future edition.*
### Tool Testing: Our Review Methodology
URL: https://www.appliedcomms.ai/tool-testing-review-methodology/
Last updated: 2025-07-01T14:00:23.000Z
The communications world is drowning in AI tool recommendations. Most appear to be written by people who've never actually used them.
Our approach is different.
## **Testing Models**
Every tool gets tested on communications work – the type we have actually done in the past, or that we are currently doing.
Tools that only work in perfect conditions aren't much use to working professionals, so we test with time limitations, looming deadlines, replicating the types of issues which will affect real-world use.
## **Our Independence Standards**
We buy our own subscriptions or use genuine free trials. No vendor demos, no special preview access that comes with implicit obligations.
When vendors offer review units or extended trials, we decline. The moment someone gives you something for free, your objectivity is compromised, even if you think it isn't.
We don't accept advertising from tool providers we review. This limits our revenue options but keeps our recommendations clean.
## **What We Measure**
- **Practical utility**: Does this save time on real tasks? How much setup is required? What's the learning curve like?
- **Cost effectiveness**: Not just the subscription price, but the total cost including time investment, training needs, and integration complexity.
- **Reliability**: Does it work consistently? How does it handle edge cases? What happens when it breaks?
- **Integration**: How well does it play with existing workflows? Does it require wholesale process changes?
## **The Bottom Line Test**
Would we recommend this to a colleague on a limited budget who's spending their own money? That's our ultimate benchmark.
We'll tell you who each tool is actually for, not just list features. A tool that's perfect for a one-person consultancy might be useless for a corporate team, and vice versa.
Expect honest verdicts, delivered in the plainest language we can muster.
### Prompt Engineering: Our Production Standards
URL: https://www.appliedcomms.ai/prompt-engineering-production-standards/
Last updated: 2025-07-01T13:57:14.000Z
Prompt engineering sounds sophisticated, but it's really just learning to ask AI systems better questions. Our approach focuses on what actually works for communications professionals.
## **Real Scenarios, Real Constraints**
We produce and refine prompts on genuine communications tasks: press releases that need writing, crisis statements that need crafting, briefing documents that need structuring.
We test with different AI models. A prompt that works beautifully with ChatGPT might fail dismally with Claude or Gemini, and the effectiveness can also vary depending on the model. We'll be specific about what we're using.
## **Our Independent Approach**
We don't take payment for featuring specific prompts or techniques. Everything we share comes from our own experimentation or reader submissions we've verified independently.
When we adapt prompts from other sources, we properly credit them – clearly explaining the changes we made and the reasons behind them.
## **What We Share**
- **Working prompts, not perfect ones**. We prefer prompts that work 80% of the time to ones that work perfectly 20% of the time.
- **Context matters**. Every prompt comes with guidance on when to use it, what to expect, and where it typically fails.
- **Version history**. We show how prompts evolve through testing. The fifth iteration is usually better than the first, and we'll explain what we learned along the way.
- **Failure modes**. What happens when these prompts go wrong? How can you spot when the AI is struggling? What are the warning signs?
## **The Reality Check**
Prompt engineering is useful, but it's not a magic solution. Most communications challenges can't be solved by finding the perfect way to ask an AI a question.
We'll always be honest about when human expertise is irreplaceable and when AI assistance genuinely helps. The best prompt is often the simplest one that accomplishes the task.
### How We Build: App Development Ground Rules
URL: https://www.appliedcomms.ai/app-building-ground-rules/
Last updated: 2025-07-01T13:52:49.000Z
When we build AI tools for communications, we're not trying to create the next unicorn startup. We're solving real problems that comms professionals face every day.
## **Our Building Philosophy**
Every app we create starts with a genuine pain point. Whether it's drafting briefing documents, checking tone consistency, or generating media lists, we only build what we'd actually use ourselves.
We use simple, accessible tools – the sort of thing that you can use with minimal or no coding knowledge. This isn't about showcasing technical prowess; it's about demonstrating concepts quickly and transparently. If something takes more than a weekend to prototype, we're probably overcomplicating it.
We're also looking to be handy with our budget: where possible, we'll use cheaper – and ideally free – tools and platforms, also to make it something which is more accessible to replicate, remix, and be inspired by.
## **Staying Independent**
We don't take money from tool providers or platforms. When we use OpenAI's API, Anthropic's Claude, or any other service, we pay standard rates like everyone else. No special access, no affiliate kickbacks.
Any code we develop is available upon request; just [get in touch](https://www.appliedcomms.ai/contact/). You can see exactly how we built something, fork it, improve it, or ignore it entirely. We're not trying to create dependencies: we're sharing what works.
## **What You Can Expect**
Honest failure reports alongside the successes. When an experiment flops (and we're sure *many* will), we'll tell you why. Even when something works brilliantly, we'll endeavour to explain the limitations we discovered.
Real usage metrics. How long did it take to build? How much did it cost to run? What wasted hours of our time? We'll share the unglamorous details.
No overselling. These are experiments and prototypes, not production-ready solutions. We'll always flag when something is rough around the edges or suitable only for specific use cases.
The goal isn't to replace human expertise, it's to augment it. See [our ethics policy](https://www.appliedcomms.ai/ethics-policy/) for more on our stance around this. Every tool we build assumes you know your job – and just need a better way to do parts of it.
### Why Applied Comms AI? (And What to Expect)
URL: https://www.appliedcomms.ai/why-applied-comms-ai-and-what-to-expect-2/
Last updated: 2025-07-24T14:36:39.000Z
**Right, let's address the elephant in every boardroom: AI.**
**If you're a communications professional, you've likely sat through at least one meeting where someone's asked, "But what about ChatGPT?" or "Should we be using AI for this?"**
If you're honest, you might have nodded knowingly whilst thinking "I haven't got the faintest clue" – or something along similar lines, but far more profane.
You're not alone. Last month, I ran a digital workshop for a multinational client (through my consultancy [Faur](https://faur.site/)), and the comms team's relationship with AI ranged from "we tried it once for a press release and it was terrible" to "we're not allowed to touch it." Sound at all familiar?
Speaking at almost the midpoint of 2025, it feels clearer than ever that AI isn't going away. However, most of what is written about it is either breathless tech evangelism or vendor marketing disguised as insight. What's missing? Honest, practical guidance from people who actually understand communications.
## Why I'm Starting This
After nearly 20 years of leading digital innovation, social media strategy, community building, communications, and content at organisations including Brunswick, Red Bull, and the BBC – working with everyone from FTSE 100 CEOs to international charities – I've seen how technology can positively transform comms when applied thoughtfully. I've also seen spectacular failures when it's not.
Now, through Faur, I work with organisations like ParalympicsGB and multinational corporate market leaders on their digital transformation. And every single one is grappling with the same question: how do we make AI actually useful for communications?
I don't have all the answers. Nobody does. But I'm here and willing to experiment, fail, learn, and share everything along the way. Think of this as your comms insider's guide to figuring out AI together.
(In time, I'd like Applied Comms AI to grow into a supportive community, but let's not get ahead of ourselves – after all, becoming overwhelmed by the possibilities is often the precise problem when it comes to AI experimentation and implementation.)
## Sign up for Applied Comms AI
Practical AI insights for forward-thinking comms leaders.
Subscribe
Email sent! Check your inbox to complete your signup.
No spam. Unsubscribe anytime.
## What You Can Expect
**In every newsletter, you'll get a mixture of sections including:**
**🧪 The Experiment**: We'll build or test something practical. Maybe it's creating an AI tool for crisis statements, or seeing if ChatGPT can actually write a decent internal comms piece. Real experiments, real results, no sugar-coating.
**🔧 Tool Test Drive**: An honest review of an AI tool from a comms perspective. What it promises vs what it delivers. Whether it's worth your time (and budget).
**💡 In Brief**: The AI news that actually matters for comms pros, minus the hype. Plus quick wins from other readers.
**✍️ Prompt of the Week**: A tested, refined AI prompt you can copy and use immediately. Because sometimes you just need something that works.
**Monthly features will include:**
- Interviews with comms leaders who are actually using AI (not just talking about it).
- Behind-the-scenes looks at our own experiments at Faur.
- Reader case studies and community insights.
## What This Isn't
This isn't another tech blog. It's not vendor marketing. It's not academic theory. And it's definitely not going to pretend AI is either the saviour or destroyer of communications.
It's pragmatic experimentation by comms professionals, for comms professionals.
## Why Now?
Because we're at an inflexion point. AI tools are finally capable enough to be genuinely useful, but not so complex that you need a computer science degree to use them. The window for competitive advantage is open, but it won't stay that way.
More importantly, I'm seeing too many talented comms professionals feeling left behind by the pace of change. That's not right. We've adapted to every other technological shift – from print to digital, from broadcast to social. We'll figure this out too.
## Our Ethics Policy
As signatories to the Global Alliance's [Responsible AI Guiding Principles for the PR and Communication Profession](https://www.globalalliancepr.org/news/2025/6/10/global-alliance-updates-responsible-ai-guiding-principles-for-the-pr-and-communication-profession), we commit to transparency first: always disclosing when content is AI-assisted and how we use it. We're also tackling AI's environmental impact by minimising unnecessary usage while we work towards better sustainability solutions. When we discover something that compromises professional standards, we won't share it, no matter how clever it seems. [Read our full ethics policy](https://www.appliedcomms.ai/ethics-policy/).
## Join the Experiment
Over the coming weeks and months, we'll explore everything from AI-powered media monitoring to automated first drafts, from sentiment analysis to crisis prediction. **Some experiments may work brilliantly. Others will likely fail and be beyond atrocious. All of them will teach us something.**
This is a community effort. I want to hear about your experiments, your failures, your questions. What's working in your organisation? What's not? What are you curious about but afraid to try?
**Here's what I need from you:**
1. **Forward this email** to one colleague who needs to see it. Seriously, just one. Someone who's curious about AI but doesn't know where to start.
2. **Hit reply** and tell me: what's your biggest AI challenge in comms right now? I'll address the most common ones in upcoming issues.
3. [**Follow me on LinkedIn**](https://www.linkedin.com/in/maclennanmichael/) where I'll be sharing highlights and starting conversations. You can find an [Applied Comms AI page on there, too](https://www.linkedin.com/showcase/appliedcommsai/).
Let's make AI work for communications, not the other way around. See you next week with our first proper experiment.
Best wishes,
Michael MacLennan
*Founder, Applied Comms AI*
**Additional Notes**
- Transparency notice: I used [Claude](https://claude.ai/) for drafting assistance and [Grammarly](https://app.grammarly.com/) for final polish, because even former subeditors need backup.
- As this is the first newsletter, I especially appreciate any and all feedback. As a former subeditor, some critical notes are the least I deserve.
- Applied Comms AI is powered by [Faur](https://faur.site/).
,