How Publicis Sapient transformed its marketing organization

If you had a dollar for each time you heard someone mention the potential of AI to transform businesses, which beach would you be sunning yourself on? Today we're talking to Teresa Barrera, CMO of Publicis Sapient, and getting the details on a real-life marketing organization transformation. Welcome to Conversations with Marta. Hi, I'm Teresa Barrera. I'm the Chief Marketing and Communications Officer for Publicis Sapient. In terms of my background, I have over 30 years of experience working technology in IT service industry in large companies such as IBM, Deloitte, Accenture, and now at Publicis Sapient. And I've spent most of my career scaling business, launching new brands and new lines of business while building high-performance teams. If I had to summarize my career in one word, I would probably say reinventing. Reinventing business, brands, teams, industries, marketing, and myself. The core of what my work, I would say, has been the last 30 years has really been about transformation and growth. All right, so I think reinvention and transformation is an excellent place to start this conversation. Thanks for being here. So you've been working on marketing transformation to help your business grow faster than the market, but like a lot of companies, you have to take into account you can't infinitely scale your headcount or your resources. So, so what does that look like? Yeah. So when you think of most companies today and how they are using AI, they tend to think of AI as a technology to drive efficiencies or reduce costs. And when you do that, as a result, you end up doing what I call random acts of AI, adding more technology to broken systems, broken workflows, and really to broken decision-making. So for us, when we started this journey, uh, 3 years ago, we wanted to use AI to help us change our ways of working and to reinvent our work. And what we learned is that very early on when we started on this journey is that simply deploying new technology was not the right approach. Basically, what did require was for us to rethink how tasks were performed, our workflows were structured, and our work is done. As I said, our transformation started about 3 years ago. And the first thing we did, I would summarize in about 4 steps, actually 4 steps. One was experimentation. We, I told the team to the entire marketing organization to experiment with AI tools and to try them, to see how they work and to be comfortable with them. At the same time, we also built our own agentic campaign management system. And in doing this, we learned basically two important things. One is that we learned by building our own campaign, the Gentec system, that first we needed to focus on tasks instead of roles. And most importantly, we needed to— the quality was more important than autonomy. We really rushed in the beginning. To autonomy and meaning make it agentic. But that process, we learned that you, we could get there, but the quality wasn't good. So if the quality isn't great, having full autonomy didn't really matter. So that was our step, first step. Then our second step after that was to really map, uh, all the tasks. So we conduct over 15 workshops where people, teams came together and basically we documented every marketing task across the marketing, all marketing functions. And in that process, we identify over 900 tasks that we, across all teams and over 600 that could be AI-led. And then what we did, it's really important, the second, it was to say, okay, then we classify which task can be done by AI, which one can be done led by a human, and which one is hybrid. And really, we concluded that about 80% of these tasks could be automated and done with AI. That's interesting. That, that actually leads right into my next question, was what did you learn about AI's performance compared to humans. Specifically, where are those places where the AI excels and where are the use cases where human intervention is necessary? Yeah, so I had this journey, right? There's probably, I would say, 4 key learnings that I would say that I would take away from it. One is that speed really only matters if you're driving growth. And, and the real value of growth is to reinvest that time into high-value work. So one of the things, part of our journey at the end, so we talked about the two steps, the other two steps was to then redefine different workflows. And redefine, like, take like a campaign. We completely redefine how a campaign goes to market by taking a sort of an AI-first approach, meaning we start look— redefining it with what can AI do and then what do we insert humans. So where, for instance, where it would take— which used to take us initially, would take us up to 20 days to— from idea to campaign, to get a campaign to market. Now it can take us up to 3 days. What would it take us about, over 50 touchpoints between different teams in this campaign process? Now it takes us probably about 9 touchpoints. So what do you want to do? The reason I say that, that's why it's really important that our goal in this transformation was, well, actually, how can we grow faster than the market? So I learned that AI does not create competitive advantage by making existing work faster. It creates competitive advantage by making possible work that was not previously done to be done now. And most, the goal for us, the goal, and I think for all transformation, the goal of transformation is not about efficiency. Efficiency. It's about the capacity to do more things that matter at greater speed. The other very important learning is that I talked a lot about tasks in, as part of our transformation, that a task is the atomic unit of transformation, not a job. And the reason it's really important to differentiate them, because a lot of the times we talk about jobs, jobs gonna be gone, automated. But a job is not equal to a task. A job is a series of tasks. And AI is very good at automating tasks. It doesn't mean it's automating roles. It's the augmentation and the reshaping of their role is what is required to accelerate this AI transformation. So for us, once we went through this transformation, we then, we decided what tasks can be automated. And then we build AI assistants to automate those tasks. Then the, like, the next step was to say, okay, now how do we reshape and redesign and reinvent each role by focusing on the work, the added value work that only humans can do? So that was really important. The, the, another important learning as well, Mike, is that context is not random. Domain experts need to own their own evals. So again, our team are building the AI assistants. We enable everybody to become a builder. And we have built over 100 assistants. The teams have built them themselves. And you can build an AI assistant within 5 minutes. To perform that task, but that task might only be performed at 80% quality. To get to the 100% quality, you gotta spend the time with the evals to giving feedback, to telling the, the, the AI assistant how that gets done. So that's really important. This episode of Conversations with MarTech is brought to you by Semrush. Meet your unfair advantage, a new Semrush uniting search and AI discovery into one powerful marketing growth engine, delivering richer data, more keywords, deeper AI insights, and prioritized next steps. With Semrush, you can turn brand visibility into coordinated action. Get a 14-day free trial and a discount. Visit martech.org/semrush-coupon or scan the QR code. You, you brought up workflows, and I think we've heard a lot about workflows. I think workflows and context as part of AI over the past almost 12 months. I'd say 6 to 12 months have gotten a lot of talk. Workflows, because for a lot of people that's sort of the second step of AI. We've learned how to use the LLMs. How do we incorporate it into the work we do? What separates a successful AI workflow from ones that fail in your experience? I think you have to rethink it. I think for us, instead of just saying, here's how I, like the example I just gave you, how we do this, this is the way we do a campaign and then just insert AI into every one of those steps. Okay. So what is going to happen? Yes, you might get faster, but you get fast, very inefficient. You, what you do is you're accelerating inefficiencies because those workflows probably were not, were broken, meaning they were not very efficient. So you have to redesign it, not just add the AI to an existing workflow, but you have to redesign it with an AI sort of first approach, meaning let's start, what can AI do? And then you redesign around that and then you add the, humans to after, to the loop where humans needed to intervene. So in some ways you have to do with an AI-first approach, but also you have to take a, you gotta, you have to take a completely reinvented, you can't just do what you are doing. And this is why to your question, I think that's why a lot of these things in the end, they really don't give you, don't drive the gains and the transformation that you actually desire. And for us, was the ultimate outcome was growth. Because it's not, and growth is not about just putting more things into market, it's putting things with quality and things that can convert. Because at the end of the day, that's what we want. Let's not just more, it's not volume, it's more, it's about, it's not driving more outcome or more outcomes, which a lot of marketing organizations still focus a lot on the output. It's the outcomes that we really needed to address. So to make sure that we reinvent that process, I think it was really important for us. Yeah, we've heard it said a thousand times that if you don't have the fundamentals in order, what AI is going to do is it's going to scale bad marketing, right? When you talked about like, yeah, it's going to help you with volume, it's going to help you produce more, but you've got to keep your eye on the quality of that more because nothing can amplify bad work like AI can. Exactly. And it would amplify bad decision-making too. That's right. That's right. Sure. We often think of AI as a technology shift because it's been one of the biggest technology shifts in our lifetime. But there are cultural and leadership aspects to all of this in order to make it work effectively, like we were just saying. So what are those shifts and how are you addressing them in your organization? I think you have to also, in parallel to this, that's what we did too. Is to build a culture that favors exploration. I call it a culture of explorers, and I tell some of the things we've done, and also a culture of builders. So one thing that we decided early on, instead of having our IT department or a group of people to go build agents and then deploy them and assistants, We decided that we're going to make everybody a builder. Every marketeer can build. And the reason why is because you know your job better than the next person or any IT person. Meaning if I let somebody that is working in PR go build these assistants, they know exactly what that assistant should be doing and they know what good looks like. So we wanted to make everybody in the team a builder, a builder of software, because I do believe in the future, you know, software is going to become very disposable. I use this analogy, it's like the Zara fashion. You use it for a season and then you throw it out, then you build more. So let's teach people how to build. So it's a culture that we intentionally created. And then the other part where I talked about exploration. How we make people into explorers. We have this analogy, we talk about there's explorers and there's villagers. And I think what AI become good as a villager, being a craftsperson, where humans become the explorers, they think ahead, they think to the future, they, you know, they look for new ways of doing things. That's what we wanted to protect humans. It's time to do that. So we try to build a culture that really embraces that. And some of the things that we do to embrace that exploration and culture is to— we create something called learning AI Thursdays, where every Thursday the entire team comes together, entire marketing organization, and shares how— what things they learn using AI, different AI tools, or, you know, what are they seeing in the market, what they experiment with, and they share with each other. We created something we call Failing Fridays. If you experiment in learning, obviously you're going to have to have a lot of failures. Most of the, the things you do, not all of them going to work. So it's really important that people feel that safety and have the space to experiment without feeling constrained or, um, by failure. So we, every Friday, the entire team comes together and they share the things that did not work, the things they tried, they didn't perform, and they learned. We have a pause for learning every first Thursday of the month. We take the entire day to pause to learn something new. We start the day with some maybe guided sessions and then Most of the day, it's self-learning for people to take the time and experiment and learn something different and new that can help them with their job. And then most importantly, I give out an award. I started this a few years ago for failure. And that means that who learn, who try something new and learn from that experience. And, you know, and so these are some of the things that you have to do to create an environment and a culture that favors progress over perfection and speed over process. Yeah, we have been conditioned for quite a while now to be consumers of technology. You buy the app or someone else builds the app and they give it to you and you use it. But like you said, who knows better how to, you know, the difference is you used to say, I know how this works, but I don't know how to build an app, right? Exactly. And now you can. Now you can because now everybody can write code and you know exactly what you need. So who's better in a position to do it? What's that saying? You know, you teach a person, if you know, make a, teach a person how to fish or you fish with a person. Yeah. So you want to teach people how to fish versus trying to fish for them. Right. Because I do believe, because remember what I said earlier, what we learn is building it is probably the easier part, but getting it to get to do it the right way and make sure the quality is great. That takes time. That could take weeks because that's where the evals come in. But the person that has to do the eval has to be the person that is really good at that activity. It is very good at that task. Yep. And usually that's not the tech person. Right. Exactly. And the great thing about AI, like, if you don't know how to do something with it, you just ask it how to do it and it tells you. So. It's like it's got its built-in instruction manual right there. Yeah. Yes. All right. Teresa Barrera, thanks for joining us on Conversations with MarTech. Thank you for having me. It was a pleasure.