There is now a clear answer to the question of how advertising agencies are using AI in 2026. Some are running it at the core of their business, generating revenue on the back of it. Others are restructuring to catch up. And the gap between those two groups is widening in ways that will directly affect every brand that spends money through a holding company agency.

The clearest illustration of this split is the divergence between Publicis Groupe and WPP in the first quarter of 2026.

WPP — the firm that Sir Martin Sorrell built into the world’s largest advertising holding company — has announced a “radical restructure.” It has fallen out of the FTSE 100 for the first time in its history. It is targeting £500 million in annual cost savings by 2028, funded by selling assets and reducing headcount across its 100,000-plus workforce. The new organisational chart collapses the old labyrinth of competing agency brands into four divisions: WPP Media, WPP Production, WPP Enterprise Solutions, and WPP Creative.

Publicis Groupe, meanwhile, reported that by March 2026, 73% of its operations and 80% of its media revenues are powered by AI — primarily through its CoreAI platform. Revenue is growing. Employment is increasing, even as the hiring rate is slower than the growth rate, because AI is doing more with fewer people. The CEO’s message is the opposite of a restructuring announcement: it is a victory lap.

These two companies operate in the same industry, serve the same type of clients, and compete for the same talent. The divergence between them is almost entirely explained by decisions made years ago about where to put the money.

The investment gap that became a performance gap

Publicis’s pivot to data and AI is not a 2026 story. It started in 2019 with the acquisition of Epsilon, the data and technology company, for $4.4 billion. At the time, the deal was widely viewed as expensive and strategically risky. Publicis was essentially betting that the future of advertising was not in creative brilliance or client relationships — it was in data infrastructure.

That bet has compounded. Epsilon gave Publicis proprietary access to first-party consumer data at a scale that none of its competitors could match. CoreAI, built on top of that data asset, is what now powers four-fifths of Publicis’s media revenues. When clients ask Publicis how it targets audiences without third-party cookies, Publicis has a real answer. When they ask how it optimises media plans in real time, it points to CoreAI. These are not features of a pitch deck; they are operational realities that WPP, for the most part, cannot replicate.

WPP’s response to the data era was different. It built a data offer called WPP Open and invested in technology partnerships. But it did not make the kind of seismic, balance-sheet-straining acquisition that Publicis made with Epsilon. The holding company structure — with dozens of semi-autonomous agencies each running their own tech stacks, P&Ls, and client relationships — made centralised AI investment extraordinarily difficult to execute.

This is the structural trap that the holding company model built. It was designed for a world where the value of an advertising group came from the diversity and independence of its agency brands. Saatchi & Saatchi needed to be creatively distinct from Ogilvy, which needed to be strategically distinct from Grey. That brand diversity was a feature. In the AI era, it became a liability, because AI investment needs to be centralised, standardised, and scaled across as many client relationships as possible to generate returns.

Publicis understood this first. It reorganised around geography, then client, then capability — a model that Sir Martin Sorrell publicly praised — rather than around agency brand independence.

What the WPP restructure actually means

WPP’s four-division structure is an attempt to do in 2026 what Publicis did in 2019: centralise. WPP Media. WPP Production. WPP Enterprise Solutions. WPP Creative. The names signal the direction: away from the cult of the agency brand, toward integrated service delivery under a single holding company umbrella.

The £500 million savings target is real. The restructuring cost — £400 million over two years — is also real. But the harder question is whether the savings create the conditions for growth, or whether they are simply a correction to a cost base that grew too large during an era when the business model could sustain it.

Forrester Research’s projection is sobering: 7.5% of agency jobs will be replaced by automation by 2030. WPP’s restructuring is partly anticipating that compression before clients demand it in the form of lower fees. The agency that restructures proactively and rebuilds around AI infrastructure is in a better position than the agency that gets restructured by client pressure — but the window for proactive restructuring is not infinite.

There is also a talent dimension that the restructuring announcement does not fully address. WPP’s new AI transformation division within Enterprise Solutions already employs more than 1,000 people. That is a meaningful investment. But Publicis’s CoreAI has been live long enough to have generated actual performance data, client adoption metrics, and proprietary training loops. WPP’s new division is competing against a product that already has years of operational advantage.

Omnicom watches from the merger integration ditch

The third major factor in the 2026 agency landscape is Omnicom, which has been largely consumed since late 2025 with the integration of its $13.5 billion acquisition of Interpublic Group — the deal that made Omnicom the world’s largest advertising holding company by revenue.

That merger came at a cost measured in people as well as money. Approximately 4,000 jobs were eliminated in the early integration phase. Three long-standing agency networks — DDB, FCB, and MullenLowe — were retired as independent brands. Omnicom has promised $1 billion in labour cost reductions by 2028, and it is targeting over $1 billion in total synergy savings through reduced duplication and technology adoption.

Omnicom’s AI platform, Omni, integrates creativity, media, data, and AI. But Omnicom is running the risk that every major acquirer runs: that the complexity of integration consumes the management attention that should be going into competitive positioning. While Publicis is iterating on CoreAI and generating revenue, Omnicom is managing the cultural and operational complexities of combining two very large, very different organisations.

This is not a criticism of the Omnicom strategy. The IPG acquisition gives Omnicom scale advantages that will compound over time. But in 2026, the short-term drag of integration is real, and Publicis is the beneficiary of a competitor that is temporarily distracted.

What brands and marketers should take from this

The agency holding company shakeup of 2026 is not simply an industry inside-baseball story. It has direct implications for every brand that buys media through a holding company agency, and for every marketer who manages those relationships.

Audit your agency’s actual AI capabilities — not the pitch. Every holding company will present AI as a core strength in the next client pitch or contract renewal. The question that matters is not “do you have AI?” but “what percentage of my media investment is actually being optimised through proprietary AI tools, and what is the performance difference versus manual planning?” Publicis can answer the first part with 80%. Ask your agency the same question and see what they say.

Understand the data asset your agency actually holds. Publicis’s advantage over competitors is not primarily about the sophistication of its AI models — it is about the quality and scale of the first-party data those models are trained on. An AI media planner is only as good as its training data. Epsilon gave Publicis a data asset that its competitors spent years trying to match and have not yet equalled. When your agency says it has “robust data capabilities,” ask specifically where the data comes from and whether it is proprietary or licensed.

Watch how agencies price AI productivity gains. One of the critical transitions happening in 2026 is the shift from time-based agency pricing to outcome-based pricing. As AI agents increasingly manage programmatic workflows, the traditional model of billing by the hour or by headcount becomes simultaneously less defensible and less accurate. Agencies that have genuinely embedded AI should be able to offer performance-linked pricing. Those that cannot are either not using AI at the depth they claim, or are still structuring their business to protect legacy revenue models.

The industries most exposed to this transition

The agency structural shakeup matters differently depending on which industry you are in. Fast-moving consumer goods brands with large, ongoing media commitments — the Unilevers, the P&Gs, the Nestlés — have the most to gain from working with an agency that has genuinely operationalised AI, because the volume of media they buy means that even small efficiency improvements at scale translate to significant cost savings or reach improvements.

Technology brands, which often run complex multinational campaigns across many markets with different regulatory environments, benefit most from agencies that have centralised data infrastructure — because that infrastructure is what enables consistent targeting and measurement across borders without a separate data team in each market.

Luxury brands, where the creative output and brand voice are the primary product, have a different calculus. The shift to AI production tools that this restructuring partly enables is a genuine threat to the bespoke craft that luxury advertising has historically demanded. The concern is not that Publicis’s CoreAI cannot produce a Chanel campaign — it is that the operational logic of an AI-optimised holding company will push toward efficiency when luxury requires inefficiency by design.

The model that is actually winning

Publicis in 2026 is not a creative agency that uses AI. It is a data and technology company that executes advertising. That reframing is the key to understanding why the divergence between it and WPP exists, and why it will likely persist.

The agencies that will lead the next decade are the ones that resolved, years ago, whether they were primarily in the business of ideas or primarily in the business of intelligence. Publicis resolved it in 2019. WPP is resolving it now. The cost of a late decision, in the agency business as in most others, is always paid in the gap between your trajectory and your competitor’s.

For brands watching this transition, the strategic lesson is simple: the agency that has already built the infrastructure is the agency that can actually use AI at scale on your business. Everything else is a roadmap. And in 2026, roadmaps are not the same as results.