The 2017 Decision That Changed Everything
At Cannes Lions in 2017, the Publicis AI strategy began with a move that looked reckless at the time. CEO Arthur Sadoun walked into the world’s most high-profile advertising festival and announced that the company would pause all awards and conference spend. The money, he said, would fund Marcel: an internal AI platform built to connect the group’s 100,000-plus staff across briefs, expertise, and open projects. The industry response was almost uniformly dismissive. Senior creatives rolled their eyes. Trade press called it a stunt. Even staff inside Publicis were among the biggest naysayers.
Seven years on, the Publicis AI strategy built around Marcel and its successor CoreAI has delivered something its critics could not have predicted: Publicis is now the largest advertising holding company in the world by market cap, having grown from $10.5 billion to more than $28 billion over five years. It grew 5.6% organically in 2025, while WPP contracted 5.4% and the peer average was, by Publicis’s own account, expected to be negative overall. It won 56% of all global new business billings tracked in 2025, recording 1,458 wins from 3,885 pitches. That is more than double Omnicom’s 656 wins in the same period.
The lesson here is not that AI wins in the abstract. The lesson is about when you build.
What Marcel Actually Did
Marcel launched formally at Viva Technology Paris in May 2018, following a Microsoft partnership announced in January of that year. At its core, it is a mobile application (Android and iOS) powered by AI trained on employee calendars, emails, and timesheets. Its primary function is matching: suggest the right colleague for a brief, surface a specialist buried three countries away, or connect available talent to an open project before that project is staffed through the usual hierarchy.
That sounds modest. The insight is that for a group the size of Publicis, with agencies spread across dozens of markets, the single biggest productivity drain is not slow creative or bad strategy. It is friction. Finding out who knows what. Getting the right team assembled. Reducing the time between a brief arriving and work beginning. Marcel attacked the operational layer directly, and did so before generative AI made such ambitions feel mainstream.
The platform’s strength compounded over time as it accumulated data. By Cannes 2024, Sadoun was telling competitors they were “scrambling” and that Publicis had a six to seven year head start. That head start was not primarily about the technology itself. It was about the data, the usage patterns, and the institutional knowledge baked into the system through years of real-world operation. That is not something you can buy off the shelf in 2024.
CoreAI: The Second Bet
In January 2024, Publicis announced CoreAI, a €300 million three-year investment in a unified AI infrastructure layer that sits beneath all its agencies. The scope is substantial: 2.3 billion consumer profiles, trillions of data points covering content, media, and performance, and 35 years of Publicis Sapient transformation data, all integrated with the Marcel assets already in place.
CoreAI spans five capability areas: insights and strategy, media planning, creative and production, software development, and operations. Half the €300 million is allocated to people. The other half is technology. That split matters because it signals something about how Publicis thinks about AI: not as a way to eliminate headcount, but as a way to make people faster and more capable. Sadoun said in March 2026 that “competitors are living in a dream about AI replacing people.” Publicis added approximately 5,800 staff in 2025 while rivals were cutting thousands.
For context: WPP shed around 9,400 positions in 2025, Dentsu cut 2,100 with 1,300 more planned, and the Omnicom-IPG merger announced 4,000 redundancies as part of its post-merger restructure. The divergence in headcount trajectories mirrors the divergence in financial performance.
Publicis AI Strategy (Marcel, CoreAI): The Results
The 2025 numbers are stark. Publicis: +5.6% organic growth, operating income up 8.1% to €2.4 billion on net revenues of €14.5 billion, market cap above €26 billion. WPP: -5.4% like-for-like, 98,655 staff at year end, a new CEO (Cindy Rose, formerly of Microsoft), and a restructuring programme targeting £500 million in annual gross cost savings. Dentsu: a record net loss of ¥327.6 billion, driven by ¥310.1 billion in goodwill impairment on international operations, dividends suspended for both FY2025 and FY2026.
The client movement data reinforces the picture. L’Oreal’s media account moved to Publicis in May 2024. Mars’s $1.7 billion media account moved to Publicis. Coca-Cola’s $700 million US media account moved to Publicis. PayPal’s global media account moved to Publicis. These were not random. They reflect clients consolidating spend with the group they believed had the best infrastructure for data-driven, AI-assisted media management.
Sadoun put it directly in 2025: “Artificial intelligence at Publicis is not a future promise, it is a reality today that is driving our growth.”
The Actual Lesson for Businesses
There is a version of the Publicis story that gets told as a simple AI narrative: they invested in AI, AI took off, and they won. That framing understates what actually happened.
When Publicis paused its awards spend in 2017, generative AI did not exist as a commercial product. ChatGPT would not launch for another five years. The investment in Marcel was made at a time when it was far from clear whether such a platform would deliver returns at all. The mockery from competitors was not entirely unreasonable. Building proprietary AI infrastructure for an advertising holding company looked like an expensive experiment with uncertain payoff.
What Publicis had was a thesis: that a connected, AI-assisted talent and knowledge network would compound in value over time. That thesis was correct, but only because they were willing to hold it for long enough. The returns did not materialise in 2018 or 2019. They materialised when the broader AI wave arrived in 2022 and 2023, at which point Publicis already had years of operational data, trained systems, and institutional capability that competitors were starting from scratch to build.
That is the pattern to study. Infrastructure built before it is obviously necessary is the only kind that creates a real competitive advantage. Infrastructure built after the moment is obvious is catch-up spending, and catch-up spending does not build moats.
For business owners and senior leaders, the relevant question is not “what AI tools should we adopt this year?” It is “what will we wish we had built three years ago, if we do not start building it now?”
The gap between Publicis and its rivals in 2025 was not created in 2025. It was created in 2017. The Publicis AI strategy, from Marcel through CoreAI, is the clearest case study available for why timing and infrastructure depth matter more than tool selection.
Avatar Studios works with businesses on AI strategy, automation, and technology adoption. If you are thinking through where to start and how to prioritise, you can find out more about what we do at avatarstudios.com.au/services.