From Watson to Lovable: How Flipside's MD is Navigating the AI Revolution

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. 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, 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.
• 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.