Publicis Groupe S.A. (PUB) Earnings Call Transcript & Summary

Good afternoon. This is the conference operator. Welcome, and thank you for joining the Publicis Groupe CoreAI Update Conference Call. [Operator Instructions] I would like to turn the conference over to Mr. Arthur Sadoun, Chairman and CEO of Publicis Groupe. Please go ahead, sir.

[Foreign Language] Everyone, I'm extremely sorry. I thought we would have kind of an informal call on Teams, and I realized that we basically have the same format than the earnings. Just to reassure you, this won't be look like the earnings. You can't see them because it seems that we don't have a camera. But we have on this call Carla Serrano, who is our Head of Strategy; we have Nigel, that is leading Sapient; we have Scott Hagedorn, that is the Head of our Data Practice; and then you have Loris Nold, that I guess you know. And I'm Arthur Sadoun. Again, we wanted to make sure that we take a bit of your time to answer any of the questions you may have had after looking at the 23-minute film. I hope you've been able to go until the end. So if you agree, what I'm going to do now, and again, I thought it would be more informal, so sorry for that. I'm just going to share a couple of thoughts through a couple of charts, And then with the team, we'll be happy to take any of your questions, of course, as what we have discussed today is pretty important for us, particularly for our people and for our clients and hopefully, for our shareholders. If I want to sum up what you just saw, I would say that there is 3 main points to take out of our presentation. The first, and by the way, trademark to Carla that is here, and will be able to tell you more, is that if you look at what is happening at the moment, I would say in the last 20 years, and I'm sure you have read that in any marketing book, it was really about how do you innovate or die. Innovation was the thing that was absolutely necessary. What we realize now, not only when we talk to clients but when we look where marketing world is going, it's really about connect or die. It's about making sure that you connect -- if you can come back on the chart before, please, it's about making sure that you connect data with data. And again, AI is nothing else than data talking to data, and we'll come back on that. But I think it is a very important notion for our investors because owning the data makes our ability to leverage AI absolutely unique. It's about connecting capabilities to capabilities and it's also about connecting people to people. And I would say this is where we started like 8 years ago when we built Marcel. Now -- and you can go to next chart, please, Julia. The only time where I'm going to bring you guys into [ China ] is to talk about connected identity. Because the thing you need to understand there is that there is no way you can connect the ecosystem if you don't use identity. You can't use cookies, you can't use third parties. You need to have identity that can connect the entire ecosystem. And with the acquisition of Lotame, and we'll come back on that later, if you have any questions, we are now in a position to see and engage with 91% of adults around the world. It was one of your very big questions for a long time now, which is okay, Epsilon is great. We are starting to build capabilities outside of the U.S. And as you know, we now have roughly 2.3 billion of people that we see through Epsilon. But there were couple of countries where we are not strong enough. Some regions like APAC, where by the way, the news that has just been announced is making a big buzz, where we needed to reinforce our data capabilities. And now we can engage with 91% around the world, which basically, we have a global view. It's very important to note that it is people, not cookies and maybe Scott will tell you more about that because as you know, [ there are differences ] that are important to understand there, We talk to people. It's updated every 5 minutes, by the way, thanks to AI and has been the case for years. It does not rely around the platform at all, which means that cookie duplication has never been the problem. And even more important, and I need to insist on that, for the last 5 years, and this is why I was saying to the team, while there was this debate do you need identity or not. I don't know if you would remember when John [ Raine ], a couple of years ago was saying, "You don't need to own the data. You can run the data." And now he came back to the conclusion that, "Yes, you need to own the data." While they were on this big debate, and we already bought Epsilon, what we have done, which is critical in our ability today to win new business and retain our clients is that we have connected this identity to the full media ecosystem, not only the publisher, where we have 90% of them that are connected and tagged to our ID but the retailers, creators and the strategic acquisition we did with creators, CRM platform [indiscernible]. And so now we are in a unique ability to not only see every client and every prospect at an individual level now, but to connect them with the full media ecosystem and to fuel what we call Intelligent Creativity. So if you look at the example that are on the team, Lola is really about how do you understand clients and prospects better. The QSR example is how can we connect all of these media ecosystem. The pharma example shows you how we can have intelligent content that reduce the number of waste but also increase business outcome. And then with auto, I think you have seen how we work with our clients, which is we stopped talking about marketing output, and we already thought in terms of business outcomes. You want to sell 10,000 cars by the end of the year? Okay. First, we need to tell you where those cars are, where are they available, who are the people that can really buy them. How do we make sure that we put the right platform for them to buy. And then you start thinking in terms of business outcome linked to our marketing investment, which is, as we said, the [ only way ]. If you can go to next chart. I will not insist enough on that. And that's something that I think for investors is very important to get, is that what makes us unique is not only that now we see 91% of the people and can engage with them. But because we own the data and because we have Sapient, we can build that in our client environment and be 100% compliant and 100% secure and being #1 in transparency. Clients does not want to rely on an external platform. And again, I don't want to be too much competitive, but that's a big difference also between us and our competition. You can always build a platform where you're going to aggregate third-party data, it's going to tell you that it's less accurate, that it makes less sense. But still, you will have a tool. The problem is that it does not belong to the client, it belongs to the agency, to the platform and relies on third party. When you own the data, you can be that build that in the current environment. And one of the big reasons why we are winning today is because the client wants to own in it because they have understood that this is part of their marketing and business transformation. We have the data, we have Sapient, where I work to bring that to everyone. The second theme, and I'm going to go faster now, Julia, please, is that -- and that's an important point is that last year, when we talked, CoreAI was a strategy. Today, it is a reality. And I don't know exactly who's on that call as can't see anyone. But you might remember, for those who were on cam last year, that we have the strategy that we are saying, remove the BS out of the AI. We thought there were too much BS around AI. By the way, there's still a lot of BS if you look at the announcement made yesterday. But we thought there were a lot of BS and we wanted to be real. And you remember, and hopefully some of you were there, we organized closed-door session to show what we mean about making a reality, thanks to CoreAI. We met with roughly 30 clients on closed doors, and out of those 30 meetings, got out like 50 pilots that we have been implementing since June. And you have seen a couple of examples and we have demoed the screen during my presentation, which are truly transforming the marketing model of our clients and truly very aligned business outcome. All of these are [indiscernible]. Last but not least, and this is something that our clients now are seeing -- Julia, please, is what we have achieved through this transformation and roughly $12 billion of acquisition over the last 10 years is really to build a category of one. That's important because, again, we have built 3 competitive advantage that, I guess, you guys need to have in hand. First, again, we are the only company except for the walled garden, the Meta, the Google that are doing that within that wall that can truly see and engage with 91% of people connected to the Internet around the world. Second, thanks to all the acquisition we have done in influencer, in marketing, thanks to our scale in media, thanks to our production capabilities, we have unique capabilities that are truly committed through the Power of One. And third, we have 25,000 engineers, mostly in Sapient, a bit in Epsilon also, that can build an ecosystem and show you into our current environment. All of this has led to what you know very well, which is our outperformance over the last 5 years and our #1 position. And it's also, and I'm sure there will be a question about that, so I can raise it now, it's also why we are so confident that again this year and despite the macro challenges of the world are encountering that we will continue to grow 2 to 3x faster than our peers, depending on how the year is going. So if you don't mind, I'm going to stop here with just one concept, which is this idea that our vision today is really to be about identity marketing and business transformation. For those that have been following us for the last 8 years, in my first call, it was Q2 2017, I guess. I came with this notion of really driving marketing and business transformation. It has become a reality that is today accelerated by the ability to do it at a personalized level, thanks to connected identity. But I'm going to stop here. I guess I'm going to pass on you all the questions we can have would be great because I think it's also an opportunity for you to know my team and the people that are truly making the success of Publicis. We are missing Dave Penski who's on a plane because he's having a pitch in a couple of hours. But I guess my friends here, Carla, Nigel and Scott and, of course, Loris, will be happy to answer any of your questions.

[Operator Instructions] The first question is from Nicolas Langlet of BNP Paribas.

I've got 3 questions, please. First of all, can you share where you are in terms of tool adoption among employees? And what are the key initiatives you are implementing to make sure that most of employees leverage efficiently the tools? And when do you expect most of the employees to be trained and up and running on the CoreAI platform? Secondly, you have mentioned the impact of CoreAI on content production. So can you remind us what's the share of net revenue you generate today in content production? And what's the midterm opportunity you see given the ramp-up of the generative AI tools? And finally, I know this is not the focus of that call, but since the sector has been weak since the WPP weak guidance, curious if you have seen any changes in clients over the past few weeks and whatever you are still comfortable with the Q1 guidance.

[Foreign Language] Nicolas, I was not expecting the last question. I don't know why, I thought everything was done. So I'll start with this one first, and then I'll move on to the share -- the tool adoption. I mean, it seems that it was a year ago now. But go back to our Q4 results, and you will see that all the negative news flow that you have heard from our competition but also, by the way, from others is exactly what we said in our call. It is true. There is a lot of headwinds outside. And this is why we came with the guidance in '25. But to be very clear, nothing has changed into what we told you, we have a very rock solid 4% that take into account everything that is happening at the moment. And we feel very confident about our guidance. Nothing has changed. It's actually, on the opposite, reinforced our feelings that we did exactly the right thing by coming with this guidance. Now if I can tell you a word about what you talked about the market reaction. I think what investors are not understanding is that it's not because the neighborhood is doing poorly that we are. It's, If you want to be honest, actually, the opposite. It's a people business, and we should not forget that. It's people creating relationship with people, creating trust and getting together into very big investments. And so when you look at the market today, you have on one side, one of our competitors that has decided to integrate more of the same in the sake of efficiencies with massive layoffs, which hopefully -- which obviously creates the kind of uncertainty for talent and for clients. And on the other, one that has to admit at this stage and the strategy that has put in place was maybe not that right, that might be time to change. So whereas you guys see kind of a weakness in the market, we see actually the validation of our strategy. And when we talk to clients, they might be sorry for competition, but they are very selfish. What they look at is a strong partner that is investing in innovation, that is ready to help them grow, that can bring them things that those others -- that people doesn't have. And this is why, again, these days are very important for us. If you look at the kind of reaction we have from our clients, which is exactly what we are expecting is after this flow of bad news, you're coming back with a very clear vision, competitive advantage that reinforces at the moment where we have less [ currencies ] in the market which will create more opportunity for us. And this is why again we are so confident for this year. Now, so did I answer your question about the forecast? Nothing has changed. Is that clear, Nicolas, yes?

Very clear.

Okay. Great. So now let's get into your question because there is a lot of great questions there. Let's start with tool adoption. And Scott, I'm going to pass on to you, but I'm going to make a point before. So there is 2 things. There is people adoption. And for those who have been there at the moment, we started with our AI platform, Marcel, in 2017, which means that everyone at Publicis, particularly during COVID and every time we do an acquisition, is onboard with what we do. We have dozens of thousands of training modules of everything we do. And CoreAI is at the center of that, and maybe Scott will tell you more. But the level of adoption, meaning embarking everyone into the journey is 100%. Just before you, we were having a townhall with more than 30,000 people, which I said on the call, I said, 30,000 people is good, but we are missing 70,000. And Loris, remind me that it was night in Asia and the news was pretty short. But still, what I'm saying here is we are extremely happy about the level of engagement and how we are embarking everyone into this journey. Now when it comes to clients, we are just starting. Let's be clear. As we told you, it's roughly 50 cases. The good news is that they are playing well which, by the way, makes us confident for this year because those clients see the interest not only to stay with us but to give us more. It's a fantastic tool in new business. And as you have seen, we had a good track record since the beginning of the year, although we don't disclose that. And to get on to the client, what we are saying is we're going to go client by client, and we're going to make sure that it becomes a reality for our people at the same time that we are implementing with our clients. But maybe, Scott, you can say a word on that, and maybe Nigel, you can also come back then on what does it mean in terms of implementation. And then Carla will get on for content production.

Hi, everybody. I'm Scott Hagedorn. As Arthur mentioned, we -- from a tool perspective, we deployed Marcel in 2017. All employees, all 100,000 employees-plus have access to Marcel. We now have CoreAI as an interface integrated into Marcel. And then if you were to look across kind of our Connected Media organization on a daily basis, now we have 200 or so global clients that are operating with our media capability, which is CoreOS. 5,000 people interact with those tools every day. So our deployment is mass. Regarding CoreAI, as Arthur mentioned, we have about 50 client pilots that are happening right now. But we've now sold CoreAI into 4 clients at a platform level, 3 of them in pharmaceuticals, one of them in CPG, and have a pretty robust pipeline of deploying. And when I say a platform sale, that is one of the sales that is in a client environment. You also asked about production and how production meets AI. We've got a kind of an interesting way or an interesting thing that we're working on, and I'll kick it over to Nigel. But we are integrating our data and identity capability into the GenAI large language models to create content, which means that we are experimenting with building studios that connect our identity capability with production, with DCO, with multivariate testing and then again, with measurement. So we see it as a kind of a transversal through line of how we can take identity through media, paid, earned, owned, shared and production. Identity becomes the foundation, Nigel?

Yes. And then just to extend that chain even further, we're also using the same set of capabilities to kind of continue and build technology in our clients' environment. So we're not just tracking how many people in our own organization are using this. We're tracking how many people in our client organizations are using this. And that is also a super important metric. But if I was to just give you a simple frame to think about this, Arthur talked about the fact that we began this journey with Marcel. When he laid out the vision for how AI was going to sort of be embedded into everything we did at Publicis, we built Marcel, and Sapient was obviously a core part of that. And then, now we've built CoreAI. And eventually, as you are seeing, we are using that same foundation to enable every single aspect of what our people are doing. And we see it very much in a way that says, it's people in terms of a physical workforce, and it's agents in terms of a digital workforce. And you put those two things together, and you get incredible outcomes. And it could be our people or our clients' people. And every one of those interactions keeps all of the knowledge institutionalized and continually learning as we solve marketing problems in one end of the spectrum and large-scale technology deployment in the other end of the spectrum.

But, Nicolas, to cut the long story short to your question and then we move on to the second one is 100% of our people are embarked into this journey. It doesn't mean that 100% will respond, by the way. Of course, there will be people that will be more interested than others. And through those 50 business cases, we roughly have 40 clients that are using one of the 4 areas that I just showed you. And to be clear, what is very interesting there is that it has to be modular. Because the idea is not to go end to end immediately. The idea is for clients to identify the areas where we can make the most progress, will be connected to their media, will be in content, will be enriching the data, will be to shifting into a model where they will truly link business outcome with their investment. And this is why we see the dynamic riding on our client is that it's now it's touching a broad base of clients with a specific project. And we go from one project to another, which is one of the reasons why, again, we are confident in our growth this year despite the macro difficulty. Carla, I'm going to pass you again on content because this is a very important point. I think it would be good if you can go with our friends on everything we have been discussing recently about why it is so important to connect content with identity. I will just make an answer with the control of Loris on the percentage, I would say that production is, what, 30% of the 20%, roughly, is it?

It's now 1/3.

Yes. So it's 33% of -- so 33% of the 20-ish percent we had in creative is production and growing roughly double digits.

Exactly.

Carla?

And Scott sort of outlined it, but part of the reason why our next-generation production model, I think everybody has been chasing the same thing of content engine, large amounts of content, driving real, real efficiency for our clients. What we're finding from our clients, now it's great, I can make thousands and thousands of pieces of content very cost-efficiently thanks to AI and GenAI and technology. But is it really driving my business? So that is why our focus is making sure that we're connecting our data and our connected identity to the creation of content, so clients know exactly what content, who to deliver it to, make sure that they're not wasting it, making sure that it's the right content to drive to the business performance. So production in a silo right now, I think we're hearing more and more that clients definitely want connectivity and assurance to business performance.

I think this is a very important point for you to understand, which is this idea that we should produce more content that is personalized, thanks to AI. It's something that, of course, is very important and something that we have been needing for a while. It's fun to see that yesterday there were an announcement about a new content engine, AI agent that WPP will invest in. We did exactly the same with Bria 1.5 years ago for exactly, and Loris can tell you more because you are part of the deal. We didn't announce it because we believe that this is a table stake, roughly. But what Carla saying is that what clients are saying is that it's great you have all of this content. It's fantastic to go and type and say, "I want this monkey to be in the red when it's flying this plane. And then I want it in blue and then I want to make sure that I can -- but at the end of the day, if you don't connect this huge amount of content with data, there is a lot of waste. And at the end of the day, what gets to the consumer makes no sense. And one of the biggest challenges in our industry today is truly to connect this content engine to identity, which is what we're starting to do pretty well. And you have seen an example in pharma. In pharma, it's very important because it's about the MLR process. It's about making sure that our content is compliant in 120 countries. Of course, the time you can gain and the quality of the work has nothing to do when you use AI and GenAI, in this case, connected to identity once again.

Just to maybe add on what you said, Arthur, about the announcement on Stability AI because you're right, it's table stake is we can't make an announcement every time we make an investment in a tool in a functionality, specifically on image generation. But we've made an investment in Bria more than 3 years ago, which is a co-investment with Intel and Getty Images. So that's one of the many, many functionality we're investing day in and day out when it comes to AI.

The next question is from Silvia Cuneo of Deutsche Bank.

Also 3 questions on my end, if possible. The first one regards one of the examples you showed on the video in relation to data privacy. This was the Lola example about the connected data. And I wanted to ask, when you show that the task to make the consumer interact on social media and your ability to track this data, does this require the consumer to explicitly allow tracking on their phone? Or are these insights derived through other means even if a user has opted out of tracking across apps, for example? So we're just wondering how you go around this data privacy point. Then second question regards the CoreAI impact on the client's budget. You highlighted how could I can save clients money and accelerate the campaign launches. So while clearly announced its efficiency, I was wondering if you anticipate that clients might reduce their overall spend or maybe reallocate in areas where there is higher return. And then just finally, around the CoreAI impact on Publicis employees and potentially driving more in-housing of certain functions at clients. Because you talked a lot about how this could impact employees towards the end of the video, and I was wondering if you could tell us about what functions in particular could see roles changing the most? And whether the fact that the clients can use the CoreAI tools, could make them bring more initiatives in-house basically?

Great. Thank you. If you don't mind, I'm going to go the reverse, I'm going to take three, then two, and then I will give one to you, Scott. And guys, feel free to add, particularly Nigel and Carla on three. So I mean, the more complex the world is, the less client wants to in-house because they need experts that can engage with the best talent and solve their problem. By the way, this is why you have seen a lot of in-housing when you have basically 2 platforms, which were Meta and Google doing 40% or 50% of our client investment. So they start thinking, okay, we could do it by ourselves but we will go for an agency. And now they realize that there are thousands of prospects, millions of influencers, thousands of retailers, CRM programs taking over. And so this in-house trend is really shifting the other way around. What client wants, and I need to insist on that, is their ability to have this technology that they own. They don't want to rent it. It's a huge cost. They want to own it for very strategic reasons. So they want to own the technology, but they want that to be operated by people that have the flexibility, the scale, the maturity and also the knowledge to make it work. So we actually think it's going to create jobs and growth opportunity, which is, by the way, what you see in our numbers. The reason why we are growing as we are, and this is getting to me to your second question, is that our abilities to build this kind of environments that are truly competitive for our clients in their environment with our resources is what generates the growth for us. And so we don't see our clients reducing their spend with us. We actually see our clients increasing their spend based on a very simple thing, which is they don't see us anymore as a communication partner that are going to run their creative media. But they see us as a transformation partner that are going to help them enrich their own first-party data in their environment, connected to the full media ecosystem, coming back to your point, where there is opportunities every day, like with influencer [indiscernible] and then fill it with the right content at the right time for the right person. All of these has an amount of fees, revenue and investment that is materially superior to what we were being paid in the past. The addressable market to which we're talking now is way bigger because we not only take money from other holding company, but from system integrators, but from the walled garden. And again, this is why we are outperforming on growth not only the holdco but the system integrators. I made a long answer. I don't know if I touch everything before we go on to the data security point. I don't know if, Nigel, Carla or Loris, you want to add something on what I just said. No?

No.

No? Okay. So maybe we go on to the Lola question on data security, which, by the way, is a very, very important question.

Yes. Both Epsilon and Lotame lead the industry in data security, privacy and compliance as basically evaluated by trade bodies specifically. Data -- what you can do with data varies very heavily market by market, and so we have essentially country-level data security and privacy policies in place. In some markets, you can do -- you can essentially connect household identity with digital identity. And there's a thing called a MAID, which is essentially a mobile identity. In some markets, all you're able to use is a MAID or a mobile identity. So what you can do on a market-by-market basis in terms of compiling that view of Lola varies. But essentially, and it's posted on Epsilon's website and on Lotame's website, but we have country-specific data privacy and security policies in place. And then we comply with kind of the opt-outs of every market as well.

If I want to -- the security point, again, is very important because our clients trust us. They are ready to do that with us and, more importantly, because we are ready to put that in their environment [ they see and trust our partner there ]. But if I want to close on your question, I would say 3 things. I would say, first, and we see that in our number, clients are actually increasing their spend with us versus reducing it. And this is why, again, despite the cuts we see in traditional advertising due to the macro, despite Sapient suffering like every other IT company -- IT consulting company at the moment of cold feet from our clients, we are still growing 4% to 5% this year, thanks to our ability to do more with our clients, okay? Second thing, because I remember that I even touched it to your point, actually, the machine is going to do the job of some of our people. And by the way, this is why we are going to improve EBITDA margin this year while continuing to invest massively in many areas, starting with our talent and, of course, starting with what we do on CoreAI, which is OpEx, not CapEx, as you know. But the point you need to take here is that at the end of the day, the reason why we are very confident that we're going to be able to bring our people into this journey and see some efficiencies coming without having to destroy too many jobs is that we have 2 things that the others don't have, which, by the way, is a big point for recruitment at the moment. First, we have the data and the technology to back everyone into this new journey. And when you are 35 to 45 years old, you still have 10 to 20 to 30 years to work, you want to be in a company where actually you're going to learn and use and superpower what you do, thanks to data and technology and AI. And the second is we are growing. And the fact that we are growing allow us to reallocate people that are doing jobs that the machine can do now and actually do it on other accounts. And that is very important because we can deliver at the same time margin improvement while making sure that we continue to bring the best talent onboard. And the last point I would make is that not only we see increase from our clients, not only we're going to be able to improve slightly our margin, but we are continuing to win market share. And again, you have seen the new business track record we had at the beginning of the year. Hopefully, we're going to continue on this kind of pace in the coming months, which is again one of the reasons why we're confident for this year. But this is making a big difference when we get into a room with the new clients.

The next question is from Adam Berlin of UBS.

Maybe a bit more technical questions. I just wanted to understand, when you talk about targeting identities through media buying, what proportion of the media billings that you, say, spent in 2024 is currently targeted at individuals using identities? Is it really small today? Is it already quite big? And how big a percentage of your billings do you think it's going to be 5 to 10 years' time? Just trying to understand that. And my second question is, when you talk about targeting identities, can you help us understand what's different when you're targeting individuals on the open Internet versus targeting them within wall garden? Can you just explain how that's different?

Yes. That's a great question. I won't be able to give you a question -- a proper answer for number one, which is how much is targeted. What I can tell you is it's -- now it's really a material part of what we do and, more importantly, it is accelerating very fast for a single reason, which is one -- for two reasons. One, we see connected TV accelerating at a furious pace. And because we have tagged at 90% of all the publisher, we are particularly accelerating in this area. So it's a lot of billions of dollars that are moving from linear TV to connected TV. And the second thing is retail media. I mean, I'm going to give you a number that I think we did not disclose, but I can give it to you is that this year, in 2024, already our clients will be spending more in retail media than -- in 2024, so last year, they spent more on retail media, which is targeted than on linear TV. So we really see a shift on both, which means that, again, I can't give you a specific number, but what I can tell you is that it's becoming material and increasing very fast. Now maybe...

Sorry. Arthur, did you say you spend more on retail media than TV? Is that what you said?

No. Sorry. So to cut the long story short, we see 2 very big things happening. First, clients shifting from linear TV to connected TV. And because we have identity, 90% of the publisher being tagged, we're able to shift billions from linear TV to connected TV, which increased materially the number of identity media that we're using. The second big shift is from, again, investment everywhere, but particularly in the linear TV to retail media, which is also identity-led targeted media that really is addressable. And when you take those two areas, you have a big part of the shift that is happening. Again, can't give you a number, but this is where we are. And maybe, Scott, you're going to be more to comment if you want, but I want to stay broad because I don't have the number in front of me, and I don't think this is the kind of number that we disclose today. Having said that, Scott, I think it would be good if you can take the question of open versus...

Yes. And I don't have the exact number either because it varies so much based on the market maturity. Some markets like India, they're not mature enough yet to evaluate that. So I can't really pan out and give you the exact global number. What I can say is it is important to think about how we use identity across the planning process. So it's not just kind of the delivery and activation of media, but it's also how we use identity throughout our planning process. And so in all the markets, we have identity built into how we plan all of our campaign activity. So it sits as a foundation for how we plan campaigns and then in the markets where the markets are mature and to Arthur's point where there's connective television is becoming mature or retail media networks are becoming mature, all of that inventory is activated through identity. So it's really a question of ultimately like the maturity of the market. But all of the people in the organization that planned campaigns have access to identity within the tools and systems to build their campaigns and audience profiles.

And just to add to that, I mean, depending on the business challenge, the media mix is going to be different, right? Sometimes it is more effective still today to use linear TV if you're going after a broad target, right? It might be more efficient and effective to do an experience. I think the Taylor Swift concerts, for example, give us a sense of that. So as you look at the media landscape, I think for some of these, the larger media mixes, you're always going to be using a certain level of sort of different activation. I think what's important, to Scott's point, though, is we are planning on an identity base before we come out with the media mix. So it might not look like it's completely personalized, and that is absolutely by design to solve that particular business problem.

Could you answer this point about the walled garden versus the open web because you have a big differentiator here, Scott. You're on mute.

Sorry, yes. One of the -- so there's a lot of things happening with the walled gardens right now. So the first thing is that we have our own persistent tagging capability. And so because of that, we're able to integrate directly with publishers and we can look at kind of planned and deliberate media and then link that back to outcomes. And so that's been a big step change in terms of what we can do. Essentially, we can help clients build their own walled gardens with their own technology. So that's been a big move in the industry. But we've also been doing a lot of work to integrate our data capability into the walled gardens with the...

And don't give -- I'm sorry, but don't give any broad on the walled garden. It's all the walled garden working.

Pardon, just all of them? Yes. So we've been integrating -- we've been integrating basically our identity capability to be able to sync it into the walled gardens for activation purposes and measurement.

And that's a very important question you were asking, Adam, because we need to make sure that we do it with publisher, but also with the walled garden under a single identity. I mean, everyone that will tell you that to only the identity and that you can play with AI or whatever it is -- so I get the point that you are different on Meta than you're on Google than you are on Amazon. And it's about making sure that you are the same unique person, that you can engage with, watch reviews on what you buy under a single identity, which is the point that Scott was making, is that if you do that well, not only we'll do a better job at targeting and delivering a good message, but you can meet your investment to business outcome. And you can know how much this person is valuable to you or is not. Nigel...

Yes. And I think actually, just on that one, Arthur, one thing to add is that what this also does is because we're building this in their environment, it's creating an asset for them which historically, when you just simply have a walled garden strategy, you're not necessarily building anything that is an asset for you as a business. Essentially, your asset is essentially the relationship you have with the walled gardens, which is a really big difference in the way and why clients are thinking about this in these two ways.

And then just one other thing that we think is special and unique is our ability to be, through identity, to look at reach and frequency across all publishers, including the walled gardens. And that can only be done through strong connected identity.

Thank you, Adam. That's fine. You've got your answer? I guess, yes.

The next question is from Christophe Cherblanc of Bernstein.

Yes. The first one is on first-party data. I was struck by the example you were giving on the auto industry, which implied you had access to the CRM system, the inventories of the clients. So do you expect over time that your model to be more embedded within the client setup? And does it mean that for clients, it will be more difficult to unplug their relationship with you, making for a more sticky model? That's the first question. And the second one is on cost. The cost of operating CoreAI, so we know about the investment you've made recently. But over time, at what stage do you expect to see some leverage on running CoreAI? And over time, do you believe it will be a business model with more operating leverage than what we've known for the last 30 years on traditional ad agencies?

Thank you. I think we're going to start with the second question, and maybe, Loris, you want to add something on that. But first of all, again, we have -- the number you need to take into consideration is how much we have been investing in technology and capabilities now. Because as we said, AI is nothing without the data and the technology. And the reason why we feel so confident and the reason why we're winning today is because we have those data and technology that we can leverage through AI. The cost of operating is pretty high at moment, almost $100 million. Now it's going to start to be financed by our clients because they are paying for the pilots, as you just heard. It's, between us, more than paying when you look at how we use that to keep clients and to win new clients. But it's still an investment. But through time, this is something that we will definitely sell as a product. But to be honest, is not that much the growth we will generate by selling CoreAI that would make a difference. It will be maybe not marginal, but not material versus our size. What will be material in terms of revenue generation is the ability through CoreAI to get into the media ecosystem of our clients and make sure that we turn the connected results to bring Sapient, to build the model with them to make sure that they go with us is for the creative, because with us, we can do fuel the system, closed-loop ecosystem that we can put around that. It will generate some efficiencies, not only because the machine will be able to ingest in some things. But on this topic, I guess, I'm going to let Loris talk before I get into your first question.

Yes, sure. Christophe, so I mean, as you know, we have been generating very significant operating leverage in the past 24 months. Because if you just look at last year, the -- normally, we've been able to deliver a margin of 18%, which is 250 basis points above the peer group. But while investing, as Arthur said, $100 million plus in the CoreAI development, $135 million in restructuring charges to upgrade our talent pool, and you fast-forward to 2025, we will maintain those investments while delivering margin improvement. Now let's be clear, CoreAI is one of the aspects when it comes to operating leverage combined with what we do at country level, combined with what we do on offshoring and overall delivery. And so that cost elasticity or that search for cost elasticity will remain and will allow us to drive that margin improvement that we have committed to you a year ago and again at the earnings.

But there is obviously some operating leverage, definitely. Coming back to your first question, I'm just going to address the stickiness of it and then I will let Scott talk about the first party data on auto. And by the way, Nigel and Carla, feel free to jump on this one because it's very important. I'm not going to tell you it's sticky. I'm just going to ask you to look at our track record for the last years now. There is a reason why we are losing less than the other. The fact that we are growing so much faster than the other is due to our new business track record, but not only. It's also mainly due to the fact that we are losing less. Why? Because the model we are building for them means that they really consider us as a partner in their transformation more than a communication partner that you can change at any step. It's particularly true for auto, for financial services, for QSR, for any kind of industry. We have the tech infrastructure and the use of data is already pretty advanced. But maybe, Scott, you can start on that, and I would like to hear Nigel and Carla also on this certain example.

Sure. The interesting thing about kind of the way you phrased it, which is it seems like you've integrated basically your capability into a client's data ecosystem, that's exactly right. Each client vertical that we're working in right now, be that pharmaceuticals, CPG, financial services, automotive, all of the client environments have unique data sets. So in the case of CPG, it could be supply chain data. In the case of automotive, it's the dealer systems, basically where the cars actually are. And those data sets don't leave a client's prem. So it's been critically important for us to plan on how we deploy in a client environment, but also how we can connect our capability to the unique data assets that each client has. And that is what really excites us because essentially we're building new marketing software applications based on the clients' environment and integrate with their unique data sets that are custom and tailored to their needs.

And then maybe just building on that, as you heard from Arthur earlier, right, you -- oftentimes, it's where connecting our data sets, but oftentimes, clients in their own environment aren't connecting their own data sets. So in so many examples, you have dealer data or in systems that are siloed in one place and marketing data that's siloed in another place and sales data at a corporate level in a different place. And by actually connecting their data sets in their environment and then connecting it to our platforms, the outcomes that we're able to deliver for them are also significantly greater. Because in many cases, they have siloed applications of AI in the organization where somebody is doing a little bit of content generation with a partnership here and somebody else is doing a little bit of identity work over there. But because those things aren't in a common workflow, because they're not connected, they can't get the outcomes that say, how do you actually sell 10,000 cars by the end of next quarter? And so effectively, what we're doing is we're connecting them in their own organization and connecting them to us, both of which, I think, makes us a super-value partner in this context.

And just -- sorry, and just in terms of a live example of this because you used the word sticky, and I don't know whether sticky is the right word. But the reality is there are no shortage of ongoing problems and hypotheses that our clients have today, especially in the economic climate we're looking at, right? So even just yesterday, one of our clients within a particular industry has some supply chain issues and we are now scenario planning what could happen given all the different scenarios that could be rolling out this year, right? So again, that data -- that shared data set and the ability to problem solve is always going to be an ongoing problem.

Just to wrap up on your question, connecting capabilities is the right thing to do for our clients and, of course, a source of growth for ourselves.

The next question is from Adi Arya of Arini.

Can you hear me okay?

Yes, very well.

I think -- sorry, there's a lot to talk about AI. I think one of the questions I had is you've obviously explained the offering you have. But when you listen to all the players, the top players doing this, so WPP, IPG, [ even as far as ] Capital, for example, everyone talks about GenAI, about being able to create a lot of content at a fraction of the time, the cost and manpower. When you speak to clients and maybe not so much with the technical, how do you differentiate? And especially with kind of the big tech partners, what -- can you give some kind of examples that I could understand about how you show them that you differentiate, like our AI tool is able to do x that no one else is able to do yet or you're developing a feature in that way.

Great question. I'm going to let Carla start there. Carla, could you -- I would like to do two things. I would like to explain on a strategy standpoint, then I will like explain how we dispatch on the data and then on the technology. So that on the two of you, if you can make it short because time is flying, but I would like to hear the two of you. I think It's an opportunity for everyone on the call to know you better.

Yes, yes. So it's -- and I think I tried to answer it before, our AI solution doesn't necessarily start from GenAI content production. I think first -- I just want to be very clear here. We start from, hey, how can we provide business solutions for our clients. And that includes our content AI production capabilities, which are -- and if I were to be fair, everybody has the same sort of content AI abilities these days and emerging GenAI tools, right? The difference for us is that because we connect those content -- that our content-making ability to connect it to identity and data, we can actually tell our clients, "Don't make 7,000 things. Make 3,000 for these growth audiences across these channels so that your business outcome will be x." Right? And I think what's been happening up to this time, there's been like this incredible thirst for making, right? Because now we can make things, as you pointed out, large amounts of volumes of content cheaply and quickly. So what we're finding from our clients is, great, now that I can do that, we get clients from our competitors that come to us and go, "I'm tired of making content that does nothing. Can you help me make it work?" And that is the difference between our strategy for AI usage for content and production versus others.

So again, hopefully, this was clear on the GenAI thing and production. Now, Scott, in terms of data capabilities, what makes ours different versus all those guys?

Yes. I think clearly that we reach now 91% of all adults connected to the Internet and that we have an identity capability that's pretty deep is what really makes us different. Identity is the through line, through all of our CoreAI capabilities. And the really important thing with our identity, if you go vertical by vertical, as I was mentioning, we can act as the data onboarder to build vertical-specific connected identity solutions. So we can still now work with best-in-breed partners for automotive data, for CPG data, for loyalty card data, you name it, and we can onboard that into our environment to create connected identity by vertical. And then that connected identity by vertical is what powers all of our AI-based planning applications and production applications. So that makes us very different than anyone else in the market.

No, just one second, to wrap up on Scott, before I come to you. This is -- this connected identity is a very important passion because we are the only one -- not even talking about the [indiscernible] that can connect our data, our identity with client data. Because if you are Walmart, for example, you don't see what your competitor is doing. If you are a CPG company, you actually can't talk directly to your consumers. So we can add what we know to what our clients know and make them stronger. Then we can connect this data to the entire media ecosystem which, again, no one can do. You can do that on Facebook, you can do that on Google, you can come to it with a single view. And third, we are able to connect this data and this ecosystem to the right content, as Carla explained it to you. And this means that we're able to have a measure number that says that for each dollar you're investing, this is the kind of business outcome you can expect. Now this is pretty sophisticated. And the last advantage we've got versus competition is that we've got Sapient that can implement that for our clients.

Yes. And I think just to talk about that architecturally and like from a platform perspective, right, we built this connected identity graph and the ability to connect that identity graph to content generation in the same ecosystem, the same workflow. And because they're built on a common platform, you can have lots of different tools doing lots of different things. And we have essentially model hubs, right? So when you talk about the big tech companies, we know this model is really good for content and images and this model is really good for text and that model is really good for code. And every one of these things can be picked dynamically on the basis of what use case we're trying to solve for the client. But because we actually have a common modular platform across that ecosystem, we're constantly building intelligence at the platform level with us and our clients in terms of how this gets delivered. Because otherwise, what you heard from Carla and Scott would be concepts. And the only way those concepts come to life is if they're grounded in technology, which is the premise of CoreAI.

It's funny just to build on that -- yes. Go on then, sorry.

Just a follow-up I wanted to ask in relation to that. The point of identity across platforms is very clear. That's exactly how I think anyone will win. So it's interesting that you -- that point you make. I think my follow-up there would just be that to what extent are you using kind of Meta, Google, the big tech products partners to do this? Because, for example, you probably have some, but when in the press I read Omnicom was awarded kind of Firefly's partner of the year, as for Capital did some kind of Meta showcase there that the [ monks flew out the gate ], et cetera. So it does seem like other people are also if not achieving the same result, a [ similar ] result with big tech. So what again, like do you not worry that they're developing these kinds of things? And again, everyone across your -- the big player cohort, you're all using big tech in different ways. And how do you see that then evolve with shareable and being able to have the best applications?

Yes. I mean that's a great question. There is not one partnership that you see on the press coming from our competitors that we don't have also. I mean it's -- the question is what do you have on the top of the other? And by the way, there are some partnerships that we are having with those guys that no one else has, and I won't disclose it on the call. But as you can imagine, as we are running on identity, we have an ability to connect our data with some of their platform that no one else can do. So the thing is a bit like the Bria thing, that we don't communicate on that because we consider that it table stakes. What really matters is what do you have that makes it really different for our clients? What can you bring to your client that a partnership with Adobe or with Google, that we can all do, will bring a difference in terms of growth for them? And that's basically what we've been discussing, which is you can have all the partnership you want with Amazon or with Google. If you don't have first-party data through identity with clean rooms where you can share that in a unique way for your clients, you are, of course, not qualified to truly bend something different. If you don't have when we talked about Intelligent Creativity based on identity, you don't have the same discussion with another being. And I can go on and on and on. So it's about the data, it's also about the capabilities. This is why we have been investing in areas where we think there is an opportunity for us to truly make a difference. Take what we did on creators. There is a reason why we are doubling down on creators. We are actually taking a very large chunk of the market, particularly when we bought the leader in [ French Code ], but also with BR Media and Lotame. Because these are kind of capabilities that we want to own in order to make sure that we can give an exclusive service and product to our clients. And last but not least, there is a point that Nigel did, which is it's great to have all of this. But if you're not able to implement it, you're missing a big part of it. And it's funny because we had a townhall just before you. And one of the question was, what is -- what are clients are the most surprised when you present CoreAI to them? And it's not the data because they know that we're winning on data, and this is why we're winning basically every pitch we go on. What they did not see is how we can build that in their environment, thanks to Sapient. This is where Sapient has such an important role to play. But I don't know, guys, if you want to add something on that before we move on to the last question?

I mean, I think the only thing, and Scott, maybe you can talk to this, is that our model makes AI affordable for our clients.

Yes. It's a very important point.

Because we own basically our own identity and because Sapient has the ability to take forward positions on cloud computing and to kind of help bring down the compute costs, we've been able to make the model work. And kind of what we've heard kind of clearly that some of the other -- some of our other competitors have struggled with making the kind of identity and data models work because they haven't been able to figure out how to make it commercially work. And because of our ownership, because of our positions with Sapient, we're able to make this work for clients. And like a CPG client can quickly get priced out if they're a low-margin business because it just becomes too expensive of a proposition to activate. And we're in a fortunate position to be able to do that here with the assets that we have.

And that's exactly why the architecture and the technology becomes critical, right? Because you can start to make the right choices about what model -- it's like essentially, if you're using OpenAI GPT 4.0 to basically ask a question about something that's super simple in your own data set, that's like using a sledgehammer to hang a small painting in the wall, right? You want a small language model to do that because it's cheaper compute, it's more efficient. And we can make those choices dynamically on the fly in the architecture within CoreAI as opposed to having to make a predetermined partnership that a client would normally do where they would sign up to CoreAI as a partner or a Microsoft or somebody, and then have to use that one-size-fits-all approach for everything, which is when you then start to get crazy compute bills, crazy cloud bills as your people start to use stuff, which is the earlier question we got.

The last question is from Julien Roch of Barclays.

My question is, you had the beginning of your journey for CoreAI because you told us that you've developed 50 pilots and you actually sold it to 4 clients.

No, no. Go on, sorry.

So you're going to develop that for your large clients, but I mean, you don't give us the stats, but the top 100 clients is about 50% of revenue. So I would think that 95% of your revenue is on the 1,000 clients? But there's obviously hundreds of thousands of advertisers in the world. And as you say, 1 of your 3 competitive advantage is Nigel's amazing army of programmers. So when -- or do you intend to launch CoreAI as like a SaaS product for smaller advertisers? And when -- well, first of all, yes or no? And then if yes, when would a shrinkwrapped version, ready to use with that intervention for [ PPC ] SaaS model sold to small clients, i.e., timing?

I'm going to leave the question to Loris, I guess. You got it is, Loris? But I want to make a correction first. That is very important. We are definitely not talking about 4 clients. We are talking roughly about 40 clients that are using some of the module now. So the 4 clients, it's definitely not the right number. We have -- what is important to take out of that is we presented this project to 30 clients in Jan. Since then, we have increased the number of presentations, and we are already roughly 40 because some have different several models that are working in a way or another on CoreAI. Am I clear on that because it's very important? We're talking roughly about 40 clients using 1 or 2, sometimes 3 different models. That's where we are. Second, on the question, Loris, do you want to...

Yes. I mean, if I understand correctly, Julien, what you're looking for is the share in our -- the share of our SaaS revenue on the data side or on the Epsilon business, which is roughly 6% to 8% of our revenues today.

Okay. My question was more if you sold CoreAI to 40 clients, CoreAI is not, at the moment, a software you can sell to your clients without anything else, where they just install the client like an Adobe Marketing Fleet or Microsoft 365, right? So my question was as you are -- as most of your revenue is, say, on 1,000 clients, but there are hundred thousands of advertisers in the world, would it make sense to develop a software that you can sell to your clients without anything else about it from them just buying the software, so a SaaS model. based on CoreAI? And you could sell that to thousands and thousands of advertisers? And if you think it's a good idea, when will Nigel be ready with that software and when can I buy it on the Microsoft App Store?

I mean, we think it could be a good idea, but this is not where we are today, to be very clear. We are not a SaaS company. Our revenue on SaaS is roughly 6% to 8% of Epsilon today. So pretty immaterial. We're talking about $150 million on average, something like that. I know you asked the question already. I'm going to be able to close with that, by the way, which is our model and, Julien, you have been following us for a while. I think we've been pretty consistent. We believe that we are uniquely positioned to actually partner our clients in their marketing and business transformation. And the beauty about AI is that CMOs and CIOs have to work together more than ever. And so what was the vision 10 years ago is a reality today. If you want to thrive in marketing, you need to firmly transform your business model. And the reason why we are winning today is that not only we can come with a combined offer between marketing and business transformation, but we can do that on identity, on data, where AI help us to leverage it in a way that create massive value for our client. And where we see the value of Publicis today, is not about packaging SaaS solutions that we can sell to smaller clients. It's actually to focus on the 1,000 clients you just talked about and moved -- while we were talking, I was receiving somewhat, moved from what is actually 50 clients today because we just checked, 2,000 in the future where we can actually help them truly transform. That's where the essence of our business model is. By the way, this is where the revenue and the margin is, even more with the kind of efficiencies we can find thanks to AI. And this is our model today. Does that make sense, Julien?

Yes. No, it makes perfect sense. I was just hoping that thanks to Nigel's amazing army of engineers, you could do both at the same time. But I guess...

So the answer is yes, we can do both at the same time. The answer is yes, Nigel is thinking about that. But the answer is also since almost 6 years now, we have always underpromised and overdelivered. So I don't want you to write a note saying that we're going to launch a SaaS business that is already working very well. One day, maybe, but let us take time because -- and maybe I'll close with this, if I may, because you're raising the right point. I mean, as you can see, there are a couple of things that we're trying to do here. First, we continue to execute and accelerate on our strategy. And the Lotame acquisition is a good example of that. We have built a model that is truly a category of one. And we are very excited about what it allows us to do. And this is where your point is very relevant, Julien, is that, yes, we might move into a SaaS business in a couple of areas where what we do can actually be amplified to smaller clients. I'm not saying it's not going to happen. But we are at the moment, and I think it was good also for you to see part of the team, where we have a talent bank that is better every day, where we have capabilities that our competitors doesn't have because it has been painful, but we have spent the necessary amount of money to get out there now. And it's really about executing our strategy. The second point, and it's very important, this is a unique moment for us where we are in the competitive landscape that gave us opportunity just to focus on innovation, on our clients and on our talent and to accelerate growth at the moment where our 2 main competitors are distracted by very different sorts of challenges, okay? And third, and this is why also we decided to do this CoreAI thing after everyone published, it was to show our strategy versus what you have seen in the last week, but also to tell you that what we said when we presented our results and we went first, as you know, which is we are living in the world full of uncertainties. But despite all of those uncertainties that our competitors are also experiencing, we are able to continue to outperform very strongly and feel extremely confident about the rock-solid 4%. That is our number. It's something that matter also because gives the calm, the serenity and the confidence to continue to win market share, to attract the best talent, again, as you will see, to continue to win some important pitches and continue to build what is really a unique model that is at the heart of our client transformation, which means that not only we grow, not only we can improve our margin, but we are creating a model that is a totally different relationship with those guys. I think I'm going to stop here. I'm going to thank you all. Loris, do you want to add something?

No, thank you.

And thank you so much for taking the time then. Thank you, and have a good day.

Thank you.

Thanks, everybody.

Thank you.

Thank you.

Ladies and gentlemen, thank you for joining. The conference is now over. You may disconnect your telephones. Thank you.