How Unilever is using AI to improve the way its marketing teams make decisions

AI has helped marketing organizations move faster. Few have figured out how to use it to improve how their teams actually make decisions.

In October, Mark Read, the former CEO of WPP, brought together enterprise leaders, AI founders and investors for prompt NYC, a new, invitation-only forum he founded this year, built for the people building AI and the marketers who implement and use it. Kathleen Dunlop, Unilever’s EVP Future of Marketing and Marketing Capabilities, joined A.Team founder and CEO Raphael Ouzan and moderator Justine Karp, VP at MediaLink, on stage to talk about how Unilever is building AI that's actually transforming how their marketing team makes decisions.

After the session, the two sat down with John Rudaizky, an ongoing strategic advisor to EY and its former CMO, to go deeper. They talked about where Unilever started, why the company no longer uses the word “pilot,” how it decides where AI is worth the investment, and how long this kind of change really takes.

Building Unilever’s “first unit of scale”

How did Unilever decide where to start?

Kathleen: It actually wasn't our chief technical officer who first briefed Raphael. It was a business leader on my team: the head of our media department in Beauty & Wellbeing in North America. He had a business problem, and he briefed Raphael to come in and help him solve it.

A.Team brought their engineers into the company, and they followed him around to his meetings. They saw the types of data he was working with and the kinds of decisions the team was making.

It really helped to have one business leader who had a very clearly defined problem. He was the leader of the team, and owner of the data.

A lot of the issues you're going to run into with these systems come down to the data. Can we connect it? Is it clean enough? Is it in the right format? The head of our media team had the authority to make those decisions. He also had the vision and persistence to see it through. And he's a great leader, so his team wanted to follow him into this new way of working.

How did Unilever avoid getting stuck in ‘pilot purgatory’?

Kathleen: We've banned the word "pilot" from Unilever. We don't want to live in that pilot purgatory anymore. What we created here with A.Team was our first unit of scale. We built it with the intention to scale, so from the start it needed to be built on our data foundation, in our environment, with the right governance and guardrails in place.

Once we had a proof point in this one team, everybody else was looking over their shoulder going, "I'd kind of like to have some of that." Can we expand this agent to do our business performance reviews once a month? Can we take it from this set of brands into another business unit? Can we take it from the US into another market? So it really starts to generate its own fuel.

What’s the biggest change in how Unilever’s marketing teams work?

Kathleen: A.Team built us an agentic system that shortened the time it takes us to make decisions on how to deploy our media. Instead of three weeks of the month spent collecting data, formatting data and understanding what the data is telling us, with only one week left to make decisions… Today we can do that as soon as the data is refreshed at the beginning of the month. Now we’ve still got three weeks to make those decisions and take action.

Choosing what to build and what to buy

What is the difference between an agentic system and an AI tool?

Raphael: There are so many AI companies and AI tools out there. Many of them are great, but often the question is: how does it all come together?

Companies need to answer how AI builds a moat that compounds and is owned by the organization – how it integrates, supports and elevates the operators and operating teams inside it. The way we do that is by building the scaffolding that harnesses the human capabilities and know-how that are unique to each organization. Our engineers and AI builders don’t sit on the edges, they come in and work with the actual team to make AI adapt to the people, not the other way around.

How did Unilever decide where to invest?

Kathleen: There are so many ‘toys’ available to us in marketing now. We've got five questions we ask ourselves at Unilever when we think about a transformation, because we don't want to fall in love with this shiny toy or that toy. We want to make sure we're actually creating value for the business.

Question 1: how are we going to create more value tomorrow than we do today?

Question 2: how is our work going to change in order to create that value?

Question 3: what is the tech that's required?

Question 4: what talent and skills do our people need to get the most out of that tech?

And finally, question 5, and the most important one: is this going to give us a competitive advantage?

If it is, then that's the one we lean into, and that's where we start. That’s how we don’t fall in love with all the shiny toys in the toy chest and make sure we're really going after the things that are going to provide value.

How should CMOs think about the stack, and about build versus buy?

Raphael: All of these playbooks are being challenged and reinvented, including the age-old question of build versus buy. We believe today the winning path isn’t to take one or the other – it’s to assemble. Assemble the unique abilities and know-how within the organization. Buy the commodities you can source from outside. And build the things that are uniquely yours, and make sure you own them.

Today it's very easy to confuse working with AI and building AI-native systems that compound. From afar they look the same. Everybody already has AI tools. It's the equivalent of lining up at the Apple Store to be the first to get the faster phone or the faster computer. You can render your video faster, but very quickly everybody else has it too.

So the change is to think less about tools and more about cycles. What are my critical, core cycles, and how do they benefit from my core expertise and knowledge, the sum of our experiences? Then build the connective tissue, the intelligence layer if you will, that gets deeply integrated into each and every workflow. And define very clearly the relationship you want your company to have between agents and humans. At Unilever that relationship is still in the making, but there's a very clear vision across the organization of building AI in the service of humans.

Building trust and getting to business impact

AI is still new. How did you build trust in the system?

Kathleen: As the first use case with A.Team, we started with a small core team that really pushed the tool, challenged it and trained it, and got to the first set of prompts we knew worked well. Then we broadened it and gave access to more people.

We didn't just say, "Okay, everybody, here's your login." We held hackathons. A.Team came in and helped us run them. The hackathons gave people hands-on practice with the tool, and they also established peer-to-peer learning, so people were sharing with each other how they were using it. In that process they were learning to challenge the tool: challenge that first answer that comes back, and don't always accept the first answer. A.Team built us a backend to the system where we can actually see how people are using it.

Why is seeing how people use the system so important?

Raphael: We’re able to see how people are learning to use this kind of resource and how their fluency with AI systems is growing, but also where the gaps are. So we can go to the specific people who need help, and run clinics as well: on the use cases that are missing, and where people are stuck and what they could do next that they never would have thought about. We’re able to track that growth very quickly, which is why the earlier you start, the more advantage you get – the roadmap becomes obvious. To Kathleen's point, at that stage it’s more about AI meeting people where they are than the other way around.

With this kind of system, people find they have more agency and more time. They can think through activations instead of spending all their time gathering data. They're reinventing themselves, and you see people truly thriving with that. The future-proofing of the organization is starting, as well as the creation of new roles.

Is this a six-month change, or a three-year journey?

Kathleen: I think it's going to take longer than most people think, but then when it gets going, it moves very fast. And I don't think it will ever end.

For the 20 years I grew up in marketing, the industry has pretty much stayed the same. The fundamentals haven't changed. But the environment we're doing it in and the tools at our disposal are changing constantly. So we're going to need people who are excited by that: people who can look outside the organization, spot something they could use, give it a try, and are excited about that constant change.

See how A.Team AI Solutions works →

A.Team is building a new model for how the world’s largest consumer brands staff and scale AI. Your teams get agentic teammates that live inside the tools they already use and take on the work that eats their time. Behind every teammate is your agentic system, built by A.Team's forward-deployed engineers inside your own environment. This intelligence layer connects your company's knowledge, records every decision, and learns from every outcome – getting sharper each cycle, for a compounding advantage you own outright. Trusted by Fortune 500 and high-growth companies, A.Team is backed by $60 million from Insight Partners and Tiger Global, recognized by the OpenAI and Claude Partner Network, and a member of Microsoft's invite-only Pegasus Program.