
The AI Alchemist
Havas' Ravi Pau On Building His Own AI Boss
The head of AI operations shares his experiences and insights on implementing AI within across the advertising hold co, including his favourite piece of work
17 November 2026
Artificial intelligence (AI) may be transforming every corner of the marketing industry, but for Ravi Pau, global AI and Innovation Lead at Havas, the conversation is less about technology itself and more about what it enables people to achieve. Tasked with helping one of the world's largest agency networks integrate AI across its operations, Pau sits at the intersection of innovation, creativity, and business transformation.
In the latest instalment of The AI Alchemist series, Pau reflects on the realities of implementing AI within a global organisation, the opportunities and challenges that come with rapid technological change, and why agencies must focus on creating meaningful value rather than simply chasing the latest advances. From rethinking workflows and unlocking new creative possibilities to preparing teams for an AI-powered future, he explains how Havas is approaching one of the most significant shifts the industry has ever faced.
Where does AI feature in your average workday?
Enough that I’ve built an AI version of my boss. It picks holes in my work before the real one gets the chance. Approvals have become suspiciously painless.
Given my job title, it would be worrying if the answer wasn’t, ‘pretty much everywhere’, wouldn’t it? I use AI every day to tear briefs apart, challenge research, pressure test ideas, and protype new things for clients. The bigger part of my job is helping everyone else use it properly. Anyone can use AI to work faster but the more interesting challenge is using it to make our work better.
What has been the greatest leadership challenge in introducing AI into a creative business like Havas?
The fastest way to lose a room full of creatives is to create a PowerPoint that tells them AI is going to make the more efficient. Nobody gets into advertising dreaming of producing twice as much work, they want to make beautiful work that delivers results for client. And they’re right to be sceptical of anything that threatens that.
The challenge is earning their trust, not forcing adoption. Less preaching and more getting our hands dirty proving that AI can make the work better, not just faster. You can’t mandate curiosity. Believe me, I’ve checked.
How do you balance speed with responsible use of AI?
Nearly two third (57 per cent) of Brits think brands using AI are cutting corners with it. So perhaps, ‘we can make it faster?’, isn’t quite the flex we think it is.
At Havas, we believe that responsible AI shouldn’t just be about what we stop people doing, but what we actively choose to do. That means making experiences more valuable for customers, giving our people the opportunity to more creative at work, protecting the rights behind the work and being mindful of AI’s environmental impact.
Our Valuable AI Experiences research shows why this matters. Across more than 21,000 AI experiences, less than 1 in 10 reached the highest standard for customer value. Yet the experiences that scored highly drove 1.6x higher brand desire. That’s a pretty compelling business argument for doing AI properly. So, when we talk about responsible AI, we need to ask a bigger question: are we using it to make things better for customers? Or just cheaper for ourselves?
Where do you think AI genuinely enhances creativity, and where should humans remain firmly in control?
AI is brilliant at making the ridiculous possible. The mad idea that would have been killed in a meeting because it was too complicated or too difficult to explain can now be brought to life in an afternoon. But just because we can make something doesn’t mean we should. AI can give us a thousand directions, but somebody still needs to know what one is worth pursuing. That’s where human imagination, taste, cultural understanding and a healthy dose of common sense comes in.
We absolutely need to own the idea and take responsibility for what we put into the world. Otherwise, we’re just making more stuff and I’m fairly sure that the internet has enough of that already.
Can you share a recent project where AI unlocked something your team couldn't have achieved as quickly or effectively otherwise?
How do you make an ad that never runs twice? For Asahi Super Dry, Havas London and Prose on Pixels worked with artist Hisham Akira Bharoocha to create an everchanging installation on Picadilly Circus.
Using our Vermeer.ai tool, the team built a bespoke system that analysed Bharoocha’s work and creative process to produce more than 400 unique designs. I love this because AI wasn’t just bolted on for a creative’s sake. Without it, creating that much work whilst keeping the artist within it would have been incredibly difficult. That’s what craft at scale should look like. And it made for a pretty spectacular bit of content too.
How do you manage client expectations around what's possible and what's still hype?
The AI industry has a remarkable talent for making everything look possible in a 90 second demo. The trouble starts when you start doing it on a rainy Tuesday afternoon with real client data, actual deadlines and two sets of legal teams.
Part of my job is keeping up with what’s new but also knowing what’s actually ready. I’m all for ambitious ideas, but sometimes the smartest idea is to start small. For a major European retailer, our team built an AI tool that improves Google Shopping product descriptions. Not particularly glamorous, but it delivered 93 per cent ROI in its first month. Getting a client to say, ‘wow’, is easy. But getting them to say, ‘that actually made us money’, is far more satisfying.
Which one AI tool could you not work without?
I’m going to give a shamelessly corporate answer and say AVA, our own AI platform at Havas. It gives me access to models like Claude, ChatGPT and Gemini in one secure place, so that I can pick the right brain for the job.
It also means that I don’t have to take out yet another subscription every time a shiny new AI tool comes along. Our finance team are particularly grateful - I think they were getting tired of my receipts.
If every ad professional could master one AI skill this year, what would you advise?
Stop obsessing over prompts and start rethinking about how you work. Look at your day: what comes in, what do you do with it, and what goes out? How much of that process actually makes sense? AI gives us an excuse to question workflows we’ve blindly followed for years, and the ability to build something better. I’ve even built myself a little wine pairing app. One of our CFOs has a wood working tool on the app store now. Because we can. That’s the exciting bit. You no longer need to be a developer to turn an idea into something that works.
Look at how you’ve always done something and ask, ‘why on earth are we still doing it like this?’
Is there a piece of work or campaign that has involved the use of AI that you would point to as the standard to beat and what would that be in your opinion?
I loved Pedigree’s Adoptable campaign, from Colenso BBDO and Nexus Studios. It’s a beautiful example of using AI to make a genuine difference, rather than just making more advertising.
The idea was wonderfully simple. Use AI to turn everyday photos of rescue dogs into professional advertising, replacing the usual dogs in Pedigree ads with real ones looking for loving homes. The ads were tailored to local shelters and as dogs found homes, new ones took their places. Within a fortnight, half the featured dogs had been adopted.
What makes this special is that the technology never overshadowed the idea. It gives thousands of overlooked dogs a better chance of being seen and adopted.
What is your one favourite prompt?
I saw a prompt doing the rounds online: ‘Based on our past conversation, what movie, book or TV character am I most like?’. Naturally I had to try it. Apparently, I’m Josh Lyman from the West Wing because I’m a ‘constructive pain in the ass’.
On a more serious note, I don’t really believe in magic prompts. We’ve made prompting sound like some mysterious art from, when really, it’s much like briefing a bright junior colleague. What are you trying to achieve? What context do they need? What should they be thinking about?



