Welcome, and thank you all for joining this session. We are here in New York with three of my fellow Directoire Plus members. Hey, Carla.
Hi.
Hi, Dave.
Hello.
Bonjour, Nigel.
Bonjour.
We have also Sam, who is our Chief Solutions Officer for North America. Sam, thank you for being with us.
Of course.
Today, we want to share with you how we harness the power of AI for our clients, for our people, but also for our operation. We will show you how we are doing it differently, thanks to our unique model that we have built over the last decade, always putting our people at the center. The least we can say is that AI was on everyone's lips in 2023, and I mean everyone, everywhere. Just coming back from Davos, it was pretty fun to see that AI could be both the problem and the solution to almost everything. Now that people are down from the mountains, I guess we will all agree that it was maybe a little bit too much. But there is no doubt that AI is going to be increasingly central in everything we do.
So we thought we should come to our clients, the financial community, and of course, our people, early in the new year, and take roughly what is gonna be one hour of your time to set the record straight on what is Publicis' vision on AI, and even more importantly, show you very concretely the real impact it will have on our business in the years to come. So let me warn you now, we are going to go deep. This is not about highlights and concept. It's about the actual work we are doing. So we'll spend a bit of time on strategy, but most of our time explaining how we are engineering it and showing you some demos about what we are concretely doing. If you are anything like me, you might have a few questions along the way.
So if you are a client, an investor, a journalist, an analyst, a partner, or of course, a member of our teams, you can scan this QR code that you should see here, and you can send us all of your question for the rest of the week. Okay, let's start now by stating the obvious. AI was not born last year with ChatGPT, at least not at Publicis. It is not something that is new for us. When you look back, we took AI very seriously and pretty early. I guess you all know this guy here, and for those that have been living in the forest in the last 10 years, his name is Satya Nadella. He's the CEO of Microsoft. This is what he said about our first partnership on AI.
What is interesting here for us is that this quote is not coming from a press release announcing one of the thousands of AI partnerships in 2023. It is from January 2018, almost exactly six years ago. Back then, Satya joined us on what was at the time, a very controversial initiative to build an AI platform called Marcel at the service of our people. The truth is, thanks to Marcel, but also thanks to Sapient and Epsilon at Publicis, we've been putting AI at the core of our model for some time now to refresh our data, create audiences, personalize content, and connect our people. Of course, in the last years, like everyone else, we have made partnership with all the usual suspects in the GenAI. We did it with a huge player like OpenAI, actually, way before the launch of ChatGPT, NVIDIA, Vertex, or Meta.
We also did smaller startups like Writer, Resemble, and ElevenLabs. On the creative front, we have also been making some truly innovative work, thanks to AI, and particularly GenAI. And to warm you up, let me show you a few examples. Okay, these are just a few examples of what we have been doing with GenAI so far in the creative space. Let me tell you one thing. Coming from a creative background, I have seen what it takes to really get to great ideas, and I don't think AI will ever replace great creative minds. Having said that, there is no doubt that GenAI will help us push the boundaries of creativity even further. Now, let's be clear. What you have just seen is only the tip of the iceberg of what AI can generate in terms of value for our clients and for ourselves.
So today, we will demonstrate that Publicis Groupe has an opportunity to sustain and actually accelerate even further its differentiation by uniquely leveraging AI, thanks to our platform organization and our unmatched data and knowledge. But to tell you where we're going and how we're going to deliver this, we need to take a step back, as getting there would not be possible without the profound transformation we have already went through. In 2017, with this team, I mean, some of you were not there yet, but with the rest of the team, we set a very simple but actually very ambitious vision: to become the indispensable partner of our clients in their transformation. To do so, we committed to shifting from a holding company to a platform company.
This shift was in service of our clients, who were looking for more and more modern capabilities like data, experience, and content, without the complexity of working with multiple disconnected specialist agencies. At the time, we made three strategic bets. First, we put data and technology at the core of our organization by restructuring Sapient with Nigel and acquiring Epsilon. There's been a lot of work, actually, a lot of cash, EUR 8 billion in acquisition, and as some of you will remember, a few bumps, too. But today, one third of our revenue is in cutting-edge data and technology capabilities, with roughly 45,000 engineers, IT consultants, and data experts. Second, we made sure that everyone can access those best-in-class capabilities by implementing a country model with a single P&L. We actually pushed even further on Maurice Lévy mantra that we love: no silos, no solos, no bozos.
And believe me, we had a couple of bozos. It was maybe the hardest thing we did, as we had to change the culture. And if we had more time, I'm sure that Carla would have a couple of story to tell us about that, no? Maybe in the Q&A we will do later. There is a lot to learn. But also, as you might remember, it has hurt our organic growth in 2018 and 2019. It was absolutely critical to our transformation. By the way, it's pretty fun to see that our competition is now starting on the same path, and I guess we are wishing them a bit of luck on this one. Last, we implemented a single operational backbone for the entire group.
Based on what we did with our shared services over the last decades, we dramatically expanded our global delivery model that was actually initiated by Sapient way before they came to Publicis, and created Marcel to put our scale at the service of our people, allowing them to share more, learn more, and actually do more. Looking back at 2017, after what were a couple of very tough years for us, our platform organization is now really paying off. We announced today that we are closing 2023 above expectation at +6.3%, significantly outperforming our industry for the fourth year in a row. Actually, our CAGR since 2019 is at +4.7%, versus 2.2% for our top four competitors, more than the double, with a gap that has been increasing last year.
Our material investment on growth has not come at the expense of our financial KPIs, which are still the best in the industry, with a record high margin at what is an expected 18% in 2023, for the second year in a row. Last but not least, by connecting Epsilon data with our Publicis Media scale, led by Dave, we have been number one in new business by far over this period, with an impressive lineup of clients. The important point we want to make here is that this profound transformation from a holding company to a platform not only allow us to outperform our industry on financial, but also extra financial KPIs. It has also positioned us today to leverage the advantages of AI better than anyone else.
What we will demonstrate now is that by infusing a layer of AI through our unique platform organization, Publicis is truly shifting to become an Intelligent System company. I know. What you're asking yourself at this moment is: What is an Intelligent System? And what will it do for Publicis and, of course, for its clients? Well... It's actually pretty simple. Becoming an Intelligent System will allow us, thanks to AI, to connect all of our enterprise knowledge. Each individual in the group will have access to everything we know at Publicis, on every expertise and every geography. To cut a long story short, we are bringing the power of everyone to the Power of One, meaning we are bringing together all the data within each and every aspect of the group under one single entity.
This entity is called the Core AI, and sits at the center of our platform organization. Core AI is trained on the data we have acquired, but also the one that we have built over these past decades. Our highest fidelity consumer data from Epsilon, our 35 years of coding experience at Sapient, our commerce shelf intelligence data from Profitero and CitrusAd, our assets from Marcel, and the marketing performance data that our 100,000 employees produce every day. To give you an idea of the breadth of our unique data sets, we know 2.3 billion profiles of people around the world. We also know hundreds of thousands of attributes that describe those people. We have an archive of millions of creative assets. Our unmatched media scale gives us line of sight into over 650 billion impression bids every single day.
We actually manage trillions of data points, capturing how marketing content and media have performed before. Last but not least, we are combining these marketing data assets with 35 years of business transformation data that are owned exclusively by Publicis Sapient. With the Core AI at the center of our platform organization, we will make sure that our trillions of data points are shareable and accessible to everyone. This will superpower our people to turbocharge our client growth. But before getting any further in our presentation and give you more concrete example of what we're gonna do, I would like to pause, actually, for a couple of minutes, to address what are maybe the two biggest challenges when talking about AI: the fact that it could destroy jobs and would reproduce bias.
At Publicis, we are committed to make AI an opportunity for our people to grow, as we already demonstrated with Marcel, and to use it with a high ethical standards. These are such important topics, that actually, Carla is gonna take us a moment to tell us more about what we are doing in those two areas. Carla, I think it's over to you. I talked too much already.
Great.
Go on.
Great. Thank you. So, like Arthur just said, it's not about super canceling your people, but super powering your people. Our strategy is focused on driving our clients' growth and innovation, which in turn, drives our own growth. Now, there are obviously a lot of operational efficiencies to be had. The effort of aggregating information, like reporting across all disciplines, starting from scratch, you know, with cumbersome RFPs, briefs, media plans. But even with those tasks, human ingenuity and original thought will be necessary to be competitive in the marketplace. As AI becomes the norm and a leveling factor, whoever can leverage direct master AI the best will win, especially at scale, across every function and capability. And believe me, that's where a platform organization is a considerable advantage. What's more, we're only beginning to understand the new jobs and roles created by an AI-led paradigm.
For example, our creative and digital experience agencies are now teeming with prompting engineers. I didn't even know what a prompting engineer was a year ago. A human-first approach to AI becomes synonymous with an ethical approach to AI. That's because we look at the ethical use of AI in the context of all people, our clients, our consumers, our employees, and broader societal impacts. For our clients, we ensure the protection of their IP through data ingestion, privacy, and confidentiality. For consumers and society, our AI ethics and responsible use, framework, and policies align AI with human values. This ensures we protect human agency, and that what we put into the world has a positive and do-no-harm impact on people and society. We also go beyond the minimum required by legislation and have established a set of ethical principles for transparency, fairness, accountability, privacy, security, and beneficence.
For example, when training AI, we use diverse and representative data to correct for bias. We ensure that we understand the limitations, protect the data that is sensitive, test and validate the outputs, all while making sure people, not machines, are in lead control. We implement these principles with a robust program, including education and awareness, policies and procedures, and compliance mechanisms, among others. As a company, we also ensure our employees have AI training to understand how to use AI responsibly and in adherence with Publicis Groupe's standard of ethics for AI. So in order to do all this, our Core AI will supercharge all our people across five key disciplines. For insight, brilliant strategy and business consultant-level intelligence will be democratized, making everyone intelligent partners. For media, media planning, buy-in, and optimization will reach new levels of accuracy and outcomes, helping our clients win at commerce faster....
For creative and production, personalized content will finally be realized at scale with efficiencies and highly relevant, desirable creative. For software, the best software and digitally enabled products will be brought to market at scale in days and weeks, not months, let alone years. And for operations, our operations and client management systems will be boosted with speed, accuracy, and efficiency. Our Core AI will ensure all our disciplines, services, and people are running, making, winning at the speed of AI. So that's how Core AI will fuel across our platform organization. I think it's back to you, Arthur.
Back to me already?
Yeah.
Okay, good. So to sum up, to become an Intelligent System company, we are building a Core AI that is a common foundation of our centralized knowledge. Core AI is at the center of our platform organization and is connected with all of our internal operation. But even more importantly, as you can see here, with our four key business pillars: data management, media, creative, and business transformation. On the top of those expertise, we are adding a unified experience that superpower every single one of our people. Like this, everyone within Publicis is a data analyst, an engineer, and an intelligent partner with all the information they need at their fingertips to accelerate growth for our clients and also for ourselves. This is our ambition for the years to come, and the good news is that this ambition is already a reality.
Now, we are going to tell you how it is becoming more and more real every day. We will start with Nigel, who will give you a true engineer view on how we have already started to build the Core AI and our plan to accelerate. Look, I'm sorry. If you are not an engineer, you might not get it all. I did not, honestly, but you can ask ChatGPT to help you out. Nigel, over to you.
Thanks, Arthur. So at Publicis, we've been working with AI for more than a decade. In that time, at Publicis Sapient, our experience has spanned across all layers of the AI stack. First, we worked with NVIDIA on designing the actual chips that were used to train ChatGPT. Next, for so many of the Fortune 500 companies, we build and integrate their data systems to be ready. Also, we've been building and deploying AI applications for ourselves and for our clients in so many industries, from machine learning-based programs for banks, helping in areas like fraud and anomaly detection, to telcos for AI-assisted customer service and promotions, travel and hospitality companies on AI-driven guest recommendations, retailers on efficient delivery and route mapping, AI pricing and merchandising content, and with agricultural companies to use vision for plant and crop care. AI capabilities have helped us rethink how software is built.
We use GenAI tools to help developers create architectures, execute agile increment plans, and increase coder efficiency by up to 40%. And critically, as I'll share in a moment, these capabilities we've built in-house provide us with the tools and assets to deliver AI faster and better. This is a level that most of our tech competitors can't match, and no holding companies are even in this conversation. As a testament to these successes, we're proud to have been recognized last year by HFS Horizons as a market leader in generative AI, in what was the first competitive review of professional services firms in the space. IDC has since also recognized our deep capabilities in driving business transformation with generative AI for our clients.
This extensive experience has given us profound insights and the context needed to understand that to make AI work, you need a strong foundational core that supports the enterprise. This core must unify the organization's data, making it not only accessible, but standardized. It requires a set of standards-based, clear ethics and protections against bias, like Carla said. It requires compartmentalization of data to ensure people or AI agents have access to only the information they should have access to, ensuring client confidentiality of IP. And most importantly, it requires a common set of AI services that empowers the whole organization to effortlessly build and innovate in this incredible space, rather than constantly reinventing the wheel, doing mundane stuff. To power our transformation to an intelligent system, we've decided to build this core for Publicis.
This is something that only we can uniquely build for ourselves, allowing us the benefit of being the users and the creators of our AI future. We call this Core AI. As Arthur said, this is an engineering view of Core AI, talking about how it will include all of our systems, Marcel, and also support new capabilities and functions as we continue to evolve. Core AI is based on the work that Publicis has been doing for almost two years on its Bodhi platform. Bodhi is a set of patterns, frameworks, tools, code libraries that accelerate our development of AI applications. Common services and APIs make development more efficient. It contains pre-built capabilities for natural language processing, computer vision, semantic search, and knowledge graph creation and management.
It contains a language model hub that manages access to multiple language models, both large language models, small language models, internally built models, and those from external partners. Core AI contains more than 100 trained micro agents, capable of autonomously performing tasks across Publicis systems and common external services. Further, Core AI is being trained on the work of our 45,000 engineers and data scientists to build models around our data assets. But let me give you a simpler view of Core AI. Fundamentally, it's got three key layers. I'm gonna start with the data layer at the bottom and work my way up. At the heart of Core AI is all of Publicis Groupe's knowledge. This is the structured and unstructured data we've created since the world turned digital.
Structured data, stored in transactional systems that are highly organized, clearly defined, searchable, and easily understood by machines, is at the heart of Publicis' data business. Our structured data spans more than 1 trillion consumer data points from Epsilon, and includes our industry-specific data, our bid stream data, asset performance data for media, pricing intelligence, as well as data from many, many more sources. Unstructured data, outside of these transactional systems, spanning documents, video images, rich media, actually makes up the vast bulk of not only Publicis' but any organization's data. It's where the huge amount of our knowledge is stored. It includes nearly 500 million lines of code we've written over 35 years at Sapient.
It includes all of our RFP and proposal information across assets we've created for our creative and media clients, like content matrices and media plans, plus all of the documents and assets within Marcel, our hub for daily interactions for all of our people sharing and collaborating every single day. Publicis' Core AI models will be trained on all of this data we uniquely possess, putting all of our knowledge at the disposal of every one of us. Now, let me ask Sam to show some of this unique data that we have powering Core AI. Sam?
What makes Core AI different than what other companies can build? All agencies already have access to marketing content and performance data, and they're definitely using this to train AI. Some agencies might even claim they see more of the marketplace, given their size, and they're gonna use this to advantage their AI. All agencies are definitely going to claim that they can see billions of IDs around the world, but they may see these IDs, they do not own them, and that is not enough. Only Publicis has Epsilon, the strongest identity graph in the United States, United Kingdom, and across France. We've also partnered to create the strongest identity offer in China, and we own Retargetly, which is the best identity graph across 11 different countries in Latin America. Publicis has the strongest signals of 2.3 billion people around the globe.
We know who they are, where they are, and because we own this data, we have the clearest line of sight into the thousands of attributes that describe these people. We know with confidence what they like, what they watch, what they read, what they buy, where and when they shop. We've taken this highest-fidelity consumer understanding, and we've connected it with our own enterprise intelligence, our line of sight into over 650 billion impression bids across video, digital, and commerce platforms every single day, and the trillions of marketing data records we have that show what has worked and what hasn't before. We have this across TV and billboards, but also digital, social platforms, and the rest of the internet. This connected marketing data asset alone is unmatched in the holding company landscape.
But there's one more thing: only Publicis has 35 years of business transformation and software expertise with Sapient. So again, why is Core AI different from what other holding companies can build? We are an organization that's expert in building AI, not just in licensing it. We've connected the best marketing data assets with the strongest business transformation experience. So with Core AI, we now have the intelligence that connects consumer understanding and marketing performance to business outcomes. Now, back to Nigel to talk about the rest of Core AI.
Thanks, Sam. The next level in Core AI is a group-wide enterprise taxonomy that our AI solutions will use in conjunction with all our knowledge assets. This taxonomy is like scaffolding, which Core AI will use to understand our business, providing a unified view of all of this data we have, standardizing terminology and semantics across various data sets, and will integrate all of these things together. The taxonomy will be critical in allowing the AI to learn of the interconnection between our assets. It allows us to use retrieval-augmented generation, or RAG techniques, better as we connect our unique data sets on the fly with large language models to provide context on specific results to users or agent interactions. Reinforcement learning, which will tighten the taxonomy in real time as our people interact with the system, and provide feedback on how information is categorized and tagged.
The standards embedded in this layer will allow us to ensure that confidential or sensitive information is never used inappropriately, and used by only those who have authorized access to that information. Finally, Core AI contains our proprietary tools for AI application development, integration, and reusable AI code, patterns, and libraries from Bodhi. This common foundation means that we can integrate the tools that we need in support of our people's work and needs seamlessly. We'll also plug in the best models and tools from our partners, including Adobe, Google, Microsoft, OpenAI, Meta, and many others, depending on the specific tasks our people have to perform. Common services and APIs will make development more efficient.
Our core integrates access to multiple language models, both large and small, internally built and from external partners, allowing both agents and end users to dynamically tap into the best models for the task at hand. Now, over to you, Arthur.
Wow! That was heavy. Sam, Nigel, thank you very much. As you can see, we spent time getting into the Core AI to make sure that you have really an engineer view out of that, because this is something serious, something that is real, and something that will make a big difference. Now, let's see how we connect into our key business pillar to deliver further growth for our clients.... In the next 20 minutes, we're gonna go faster now, Dave will tell you how we will drive media effectiveness. Carla, how it will allow us to drive personalization at scale. Nigel, how it will allow us to deliver software even faster. And finally, Carla will come back to explain how it will boost our operation for our people. I love when you talk about operation, Carla. It's gonna be awesome.
Let's start with Sam again, who will share how the Core AI can deliver better insights and truly transform our people into intelligence partners. Come on, Sam.
Yeah. Thank you, Arthur. So what do we mean by an intelligence partner? Imagine if your marketing agency always had all of the data needed to answer your exact question completely and quickly. They could pinpoint the sources of business volume, anticipate consumer demand, calibrate your plans to your own business signals, and optimize your investment to your outcomes. Imagine now that your agency starts to anticipate your questions, giving you intelligence before you even asked. They'd always be learning and always optimizing. An intelligence partner is an analyst, a strategist, and a marketer all at once. With Core AI powering our insight capabilities, we are supercharging all of our teams to be intelligence partners. We will be faster. Every Publicis analyst will deliver the highest fidelity insights, processing trillions of business and marketing data points quicker than before. We will be more accurate.
Every Publicis strategist will convert insights into the highest propensity audiences and platforms that will drive the growth we're looking for, with the strongest confidence to your business outcomes. We will do things that we've never done before. Every Publicis planner will create infinite amounts of personalized content, optimized for production and distribution at scale. Let's start with a concrete example, looking at automotive, an industry that we all know is going through historic changes. With high expenses, low supply, cumbersome dealer networks, and tight competition, our marketing needs to move cars, and we need to do so quickly. In this example, our client's goal is to sell 40,000 electric vehicle SUVs this year. The first task of any team is to understand category and brand headwinds and opportunities. Our strategy teams upload briefs for Core AI to process and analyze.
Now, we immediately highlight what is most important in that brief, automatically generating relevant data and research, like competitor moves, market trends, and customer motivations, all derived from the highest fidelity assets and marketing data. This work used to take us weeks, and now it's only taking us seconds. Instead of having teams of data scientists working to clean, and then to harmonize, and then to analyze data over the course of weeks, our strategists immediately respond with clear analyses of the opportunity and way forward. This is the power of Core AI, processing trillions of data points to deliver business-ready intelligence across the company. We are also strengthening the accuracy of our work with Core AI. Our teams need to precisely identify sources of volume and demand for this SUV. Now, thousands of Publicis strategists can work like they are data scientists.
They are instantaneously building rich consumer profiles with Epsilon's leading consumer data, enriched with auto buyer and vehicle detail. With an accurate and precise picture into consumers' on and offline behaviors, their attitudes, motivations, and their life stage events, every single one of our strategies is built with the strongest data foundation. It's also every marketer's dream to respond to business challenges as quickly as they come. What once felt like this dream, we can now make a reality. With Core AI, we can anticipate, replan, and optimize now in real time. In our auto example, our client's business is now forecasting low supply, and we are immediately signaled of upcoming vehicle shortages in key markets. We need to redraw our growth segments and do so quickly, focusing on the markets that will have cars on lots next month.
Our new consumer audience strategy and personalized plans are ready in just minutes. This is what it means to be an intelligence partner. Now, Dave will show us how Core AI will help drive media effectiveness.
Thanks, Sam. We've all seen how Core AI accelerates our ability to pinpoint consumers that will drive business volume. So how does Core AI advance media strategy and activation? We will be faster. Every Publicis platform expert will launch campaigns immediately with operational excellence, ensuring we're in market before our competitors. We will be more accurate. Every Publicis analyst will intuitively optimize campaigns to outcomes quickly, holding marketing budgets accountable to business outcomes. We will do things that have never been done before. Every Publicis planner will now deliver limitless planning, endless versions of personalized media plans, recalibrated for business signals and performance outcomes. Let's go to the demo. Armed with Core AI, our planners begin to calibrate a media investment strategy to deliver the vehicle sales goals in the client brief.
Our planner uses a chat function to enter a series of prompts, asking pointed questions like: How much investment is needed to move 40,000 SUVs this year? Or, where will we have cars on a lot next month? With the connected intelligence of Core AI, we process trillions of data points that compare our new audience strategy to historical performance and marketplace opportunities that exist. Our planner instantly generates media plans across paid and owned channels. We've accelerated tactical weeks of planning into mere moments with Core AI, and we don't have to sacrifice accuracy for this speed.... With our investment strategy now set, we set campaigns live with Core AI. We traffic and launch assets through automated QA and tagging, cutting weeks of manual work down to minutes. We prevent errors and time loss, getting us into the market faster than all of our competitors.
With Core AI, we'll be more accurate. With the campaign now live, we use Core AI to guide optimizations. Our analysts immediately process trillions of rows of business and media performance data, calibrating investment decisions and bid strategies to meaningful outcomes every moment of the campaign. We can do things that have never been done before. Now, Core AI's intelligence shares an alert mid-campaign that shows information on national and state-level offers for electric vehicle buyers. With Core AI, we immediately tailor new plans and messages in real time to launch search, social, and email campaigns with a new personalized offer. We coordinate these adjustments to paid and owned channels simultaneously, wasting no budget or time as we drive qualified customers to dealers which have inventory in stock. Throughout this campaign, Core AI is continuously improving.
We calculate, process, and respond immediately to pacing signal tools to drive better outcomes. More importantly, we maintain complete confidentiality at every step. With secure sandboxes, controlled access, and clear audit protocols, we ensure complete compliance for all of our clients. In sum, the real business value here is our connected intelligence. All audience and campaign performance data is pushed out to every member of the agency team, wherever they sit around the globe, as well as the client's brand and marketing organizations. We have the data we need in decision time, and we're advancing our intelligence for the future. Now on to Carla, who will tell us more about creating and producing.
Thanks. Thanks, Dave. Now let's see how we can leverage all of this intelligence on the sources of business volume to power personalization at scale with the speed and precision of AI. What does this really mean for our clients? Well, guess what? We're gonna be faster. Our creative teams will collaborate more seamlessly, accelerating the creative process with brand-safe AI of GenCanvas. We're gonna be more precise. Our producers will rapidly produce content personalized for micro audiences and fit for platforms using AI-assisted production tools. We're gonna do things we've never done before. Our content strategists will test and optimize campaign assets in real time, rooted in outcomes performance data. Let's take a look at the CPG industry and see how we can leverage all of this intelligence on the sources of business volume to power personalization at scale with the speed and precision of AI.
With rapidly shifting consumer preferences, CPG companies have to not only produce engaging content, they need to produce it at a massive scale and for very localized audiences, ultimately targeting an audience of individuals. In this brief example, we'll be looking at a heritage cookie brand, guess which one, that's looking to reengage their older audience through a packaging rebrand and a new campaign launch, all while protecting their core target. To answer this brief, our creative and production teams will use our Core AI and creative capabilities to unlock new understanding, explore and develop diverse concepts, rapidly creating personalized content that will resonate with consumers, driving consideration and purchase. Let's enter the workspace of a creative director at Publicis as he's working on a campaign for our client's heritage cookie brand.
Within the Project tab, he has all that he needs at his fingertips: the creative briefs, audience insights, and big ideas for the brand's relaunch. Now, with Core AI, he has direct access to our growth and core audience segment insights, understanding their behaviors and interests as he develops concepts. With our Core AI and creative capabilities, he has all of the incredible visuals, concepts, and taglines from yesterday, as well as access to GenAI-enabled tools to accelerate his creativity. Within GenCanvas, instead of sending versions back and forth as they do today, our creative teams collaborate in real time. They use image, tone, and video generators to breathe life into brand personas and iterate brand and product content that will resonate most with our audiences.
Next, with these new tools, production leads can leverage AI to automatically create content matrices based on tactical plans, as well as past client, campaign, and industry knowledge for efficient planning for production. All the iterations needed to launch a campaign are laid out, reducing days of work to seconds. Now, instead of painstakingly recreating templates and layouts, the production team lets AI do the heavy lifting. This frees up time to dive headfirst into crafting innovative and eye-catching campaigns, all while maintaining full creative control and customization. AI-designed templating tools empower them to design with agility, drawing inspiration from historical insights and audience data, so they can focus on creating resonating and impactful work.
Once satisfied with master assets, the production team can go from one asset to 1,000 versions in minutes instead of weeks, automating asset variations for different platforms, languages, contexts, and specs, powered by market-leading and brand-compliant adaptation and versioning tools. This is something we could only dream about in the past. All the while, they are still in the driver's seat to review, modify, and enhance assets for target audiences, translations, or even replacements of the product with a best fit recommendation for audience and channel. With all these assets watermarked, reviewed, and tagged, the team is empowered to manage the scale of content like never before. With AI, our teams will continuously test, learn, and optimize campaign assets, building intelligence for campaign optimization and future planning.
... Now, that was all about the creative and production engine. Now let me show you how AI affects super shoots and development for personalization in the real world. This year, after a record year in 2023, we decided to personally thank each of our 100,000 people for their incredible effort. Personalization at scale inside. Let's take a look at how we did it.
I'm ready when you are.
At the end of every year, the senior leadership at Publicis Groupe goes all out for Wishes, a gigantic celebratory thank-you video made for the employees who helped make the past twelve months so incredible. But what do you do in 2023? An incredible year of firsts, and also the year that the AI revolution arrives, promising to change filmmaking and creativity forever.
Well, you use AI to make the best Wishes ever.
What is this?
It's me, your AI twin.
But I don't have a twin... an AI. I don't trust AI.
As it turns out, AI is an incredible tool when guided by human creativity, and it helped us create the best, most elaborate project in the history of thanking people. A technology that allowed the Directoire Plus, Carla-
Ta-da!
Dave-
No AI horses were harmed in the making of this video.
Agathe-
Bravo!
Nigel-
I think the AI me is gonna throw up.
Along with Maurice-
Here I am.
... to finally reach their true potential in every reality.
But we couldn't fit all these jokes into one Wishes video, and because our AI selves were able to cross time and space, they appear in 100,000 individual films. That's right, one for every single employee who works at Publicis Groupe.
To make each piece special, we tapped into the vast amounts of data available on Marcel to create a totally personalized video for each recipient, celebrating their interests and their passions.
Yuri, Marcel tells me you like dancing, right?
Eva, Marcel tells us you like sports, right?
So how did we pull it all off? Well, we essentially trained image and video generation models using basically every currently available AI tool in an interconnected workflow to create digital facsimiles of multiple characters, preserving their likeness across different shot types, all built on the backbone of a trained GPT script model that is capable of translating lines into 30 different languages.
Basically, creative filmmaking at a massive scale, forever changing how we think about AI's role in creativity and the capabilities of Publicis Groupe. You could say we've perfected AI.
But in the end, Wishes 2023 wasn't about technology, it was about making every single person feel seen.
Laura.
Tony.
Eva.
Yuri.
Gabby.
Which, after the amazing year we had, felt like the most important thing we could do.
Because really, what's the purpose of technology if it's not in service of people?
Wow! I'm kind of starting to like AI. I couldn't have said it better myself.
We wish you a happy and healthy new year.
That was super cool, right? I hope you find that as amazing as our people did. Now, let's go back to the technology part of our business. Let's see how we use Core AI to boost our digital transformation operations and deliver software even faster. Today, engineers make up close to 40% of Publicis Groupe employees, building software that powers cutting-edge digital transformations for our clients and internal teams. We're gonna supercharge these teams, help them innovate and build using Core AI and our software capabilities. We'll be faster. Our engineers will be able to move from idea to live at speed, with AI tools helping them code, work upstream in the engineering process, creating new architectures, test cases, and documentation, with AI ensuring that they don't sacrifice quality or security for speed. We'll be more accurate.
Our teams will build better personalized digital experiences linked to behavioral insights and audience data, ensuring higher engagement and conversion rates. But most importantly, we'll do things we've never been able to do before. AI will work with our engineers like a virtual team member, helping them architect digital solutions and learn to adopt new technologies, allowing our people to grow and adapt in ways that was simply not possible before at scale. Let me show you how. Core AI and our software capabilities will help us develop better software, using the case of a travel client looking to implement an integrated loyalty program into a new channel. As the engineer opens the system, he finds stories that a scrum master working with Core AI has assigned to him based on his profile and skills.
Now, instead of having to read through pages and pages of documentation, for those of you who code, the engineer now starts chatting with Core AI. Through simple conversational interface, he works with the system to review the requirements and validate the estimates of each story. Core AI gives the engineer a virtual twin, who can help develop a plan and address and set due dates for each story. The AI suggests that he starts on the architecture story based on the potential dependency graph of all the stories assigned to him. A Core AI agent generates a relevant sample architecture to start from, based on similar work we've done in the past, and the architecture standards of this project. Something like this has just never been possible. It also highlights appropriate starter kits and accelerators specific to the platform and requirements of the story.
We've already deployed an alpha of this capability, and it can reduce the time to do these tasks by up to 40%. Core AI now creates the initial documents and helps schedule a review board to review the architecture, all from within the chat, instead of having to search Outlook calendars for available time. The architecture story is marked closed, with real-time tracking data added to Jira. He moves on to the next story, and Core AI gives the engineer a better insight into the outcomes needed to implement the service for a loyalty points-based reservation systems. Ready to code, he uses Core AI to plug in his IDE to implement the story.
Core AI explains the broad concepts embedded in the code, and the engineer leverages Core AI for suggestions, best practices, reusable code, enterprise common libraries, accelerating the development time, but also ensuring the code meets quality standards. He creates cloud deployment scripts using AI and leverages plugins to deploy these features to the cloud. With support from Core AI, he's ahead of schedule, and the due dates and all the other dependent stories are automatically updated, giving the broader team a clear view of his timeline. So you see how Core AI and our software capabilities make our engineers faster, but much of our vision is to truly change how engineers work in ways that were never possible before. Finally, let me hand back to Carla, who's gonna show us how we'll use Core AI to boost operations and client management. Carla?
Thanks, Nigel. A little bit about operations. I wanna talk a little about how AI will help all 100,000 of us at Publicis Groupe stay more productive and connected. We expect that its impact will be far-reaching in how we run our business. We're gonna be faster by simplifying our internal processes and letting Core AI handle the tedious and routine back-office tasks. All our talent will save valuable time to focus on high-impact work. We're gonna be more accurate. I'll take one example here of how our sales and client leads will be able to understand their client better and generate winning proposals faster by using Core AI to leverage our collective knowledge. And to keep confidentiality, we'll show it being used for an imaginary theme park provider, Wonderful Parks.
As for what's never been done before, trust me, once we find it, we'll deploy it in service of efficiency, effectiveness, and performance. Let's take a look at this demo. As soon as our group client lead receives an RFP, she now has instant access to insights and knowledge to plan a better response. She uploads the RFP into the system, and powered by our insight capabilities, she receives market analysis, competitor comparisons, and other insights to help her get smart on the client. Now, using a simple chat interface, she can learn about prior work we've done with the client across the group and access summaries of recent projects, leveraging data from Core AI. She can also dive deeper into market and expert insights, getting smarter with the help of AI and learning about recent events and announcements from Wonderful Parks.
No longer relying on just who she knows, Core AI now assists her in assembling a team, and the group client lead is simply able to ask the system to identify potential team members based on expertise, availability, flag any potential client conflicts, all without having to email dozens of people. Now, she's not slowed down by having to wait for responses from everyone. What used to take a week is reduced to minutes, as the system helps her easily generate an initial draft response with pre-populated answers to repetitive parts of RFPs and relevant Publicis case studies, allowing her to focus on applying her unique insights that will make the proposal magic. Being able to create the bulk of the RFP response at the snap of a finger is something we have all dreamed of before, but now it will be a reality.
With automated assistance, she can manage all the details of the RFP response through the interface, including exploring pricing guidance and tailoring content to match what she thinks will resonate most with the client. The system will even pre-fill standard RFP questionnaires. All she has to do is review and approve. With Core AI, she has tapped into the knowledge of all of Publicis Groupe before she even convenes her proposal team and gets them ready to nail the pitch. Back to you, Arthur.
Carla, you have massively improved your operational skills.
I know.
Awesome. Thank you, guys. I think we are roughly on time. We are actually very on time, if I look at timing. We have gone through Core AI and showed you how the potential it has to create for our client growth when you connect it with our core capabilities. But at Publicis, we know that we have the greatest impact when we connect all of these capabilities that you see here together at the service of our client transformation. Nigel will now briefly talk to you about what we call the unified experience that I mentioned earlier, that will connect these capabilities in a seamless and collaborative way for our teams and for our clients to accelerate even further the power of one. Nigel, over to you.
Great. Thanks, Arthur. So let's take an example. A strategist will now better understand a customer audience through an agent responsible for creating the voice of a customer, that the agent itself is relying on micro-agents that can autonomously cull through social media and community site posts to gather the right insights. Our media planners will build on that audience information and use it to create AI-assisted media campaigns, while in parallel, our creatives are building data-informed, personalized assets and optimizing on the basis of media performance that flows back into the system. Our teams will feel seamless as we make it even easier to create incredible outputs for client in our Power of One model. Our teams will be able to move fluidly between these capabilities and maintain a consistent understanding of audiences, assets, campaigns, thanks to this unified experience. So what happens next?
A team led by Publicis Sapient, Epsilon, and many others across the group are already at work building the initial proof of concepts, AI models, and MVPs. Across the first half of this year, that work will begin to roll out to select users for testing and feedback, and that feedback will drive iterative releases, which will add more functionality. At Viva Technology, we will assure the first MVPs in the second half of 2024. Our plan is to start the group-wide rollout of this functionality, building towards a vision that we have shared with you here today... We'll continue to iterate, listening to feedback and ensuring we pair the best tools available in the market with our own capabilities. Let's be clear, this isn't something that's gonna happen overnight.
It's a powerful proposition that will fundamentally change the way we do business every single day for our people and for our clients. Over to you, Arthur.
Thank you, Nigel. Okay, now we are moving to the money charts. For our investors that have been looking since the beginning, you went through the strategy, you went through the engineering part, that was pretty heavy. We went through five cases, and now let me tell you a word about how we plan to invest in the future. We will actually invest EUR 300 million over the next three years, all out of our P&L. For 2024 alone, we anticipate an investment of EUR 100 million that will be split 50/50. Half on people through upskilling, training, and also recruitment. Half on technology, through licenses, IT software, and cloud infrastructure. This EUR 300 million plan will be fully financed by efficiencies generated by the group agile structure that you know, but also by Marcel.
This means that while increasing our competitiveness, this investment will have no dilutive impact on our operating margin in 2024, and will be slightly accretive from 2025. Voilà! We are coming at the end of this presentation. Our journey from a holding company to a platform has not been easy, but it definitely paid off. As you have seen again today, with our organic growth in 2023, outperforming our industry for the fourth year in a row. This profound transformation now positions us to push the boundaries even further by putting a core AI at the heart of our organization and truly becoming an intelligent system company. Leveraging AI will definitely make our people more efficient and increase our productivity. But this is not what matters the most.
What matters the most is that by shifting to an intelligent system company, it will allow our people to do things tomorrow that no one can do today, as you have seen in the example that Carla, Sam, Dave, and Nigel shared with you. Let's be clear, we will be the only ones able to do those things tomorrow. Thanks to the transformation we went through, no one has the breadth, the accuracy, and the reach of our proprietary data. No one has been dramatically transforming their model to get to a single structure, and no one made the kind of investment we did with Publicis Sapient and Epsilon to put at the heart of our model 45,000 engineers, consultants, and data analysts. Just because we are the only one who can do it, doesn't mean that we will get it all right.
Actually, AI needs experimentation, and some of the things we just showed you will be wrong at first, but we will learn, and we will fix them. By starting first with a unique model built over the last decade, we are confident that we can leapfrog the market once again and continue not only to deliver the best margin in the industry, but also to keep growing faster than our peers in the year to come. As you may know, at least if you're French, Publicis was founded in 1926 by Marcel Bleustein-Blanchet. Marcel used to say, "In reality, it is not creativity that evolves. It stays the same. It is everything else that grows around it." This is exactly what we have done in the last decades, mostly under the leadership of Maurice Lévy. We have grown and transformed everything around creativity for almost 100 years now.
As we are about to enter into our next century in 2026, hopefully, you have seen today that our commitment to lead the change in our industry, for the benefit of our people and our clients, is stronger than ever. Thank you for listening, and see you in May at Viva Technology. Merci.