AI is reshaping Omnicom’s workflow. Its revenue model may be next

Which explains why Paolo Yuvienco, Omnicom’s chief technology officer, was on the call.

He mapped out the group’s move from basic generative AI use cases — think copywriting and content ideation — to something more ambitious: a proprietary network of AI agents embedded across the organization. But these aren’t just task bots, said Yuvienco on the call. They work in tandem, powered by Omniocom’s Omni data platform and tuned through its own institutional knowledge. The agents simulate focus groups, tweak campaigns in real time and surface signals from shifting sales trends. It’s less about automation for efficiency’s sake, and more about building a moat, the CTO continued.

“Every discipline and all the teams across Omnicom now have the capability to drive deeper intelligence and a deeper understanding into every part of the work that they’re doing for clients,” said Yuvienco. 

That includes strategy and creative teams across Omnicom’s agencies, which are now working with synthetic audience agents grounded in Omni data. The result: synthetic focus groups for ideation, personalized content development, pre-launch testing and campaign scoring. Or as Yuvienco put it: not automation for automation’s sake — AI as creative infrastructure.

But for all the talk of scale and speed, there’s a subtext hard to ignore: the more AI becomes infrastructure, the fewer people are needed to build around it.

That tension isn’t lost on Omnicom’s leadership. As CEO John Wren put it on the call, the financial impact of AI is still “a book yet to be written”. So far, the gains are largely internal — teams working faster, smarter, more efficiently. But the real inflection point hinges on client adoption. Unlike self-serve platforms that cater to smaller businesses, holdcos like Omnicom are built around complex, global enterprise. Those clients move slower. Their willingness to pay for AI-powered services will take time to mature. 

Wren offered a telling example: he recalled an ad generated in minutes using AU — featuring a cat in a hat — that was promptly rejected by a client’s legal team. The issue wasn’t technical but compliance: “It’s illegal to put a hat on a cat,” he quipped. The line was tongue-in-cheek but the point was serious: large enterprises have entrenched approval processes that can stall even the most advanced tools. 

Plus, there’s a looming question of cost. As Wren noted, the costs of the infrastructure needed to scale AI — compute, storage, systems, integration — hasn’t hit the balance sheet yet. Ultimately, what clients buy into won’t just come down to what agencies can build, but whether they’re able to stomach the true costs of deploying the fanciest product on the market, versus sticking with what already works, remains to be seen. 

For now, Omicom’s strategy is clear: get the tools in place, equip teams to use them and let innovation emerge from the inside out. The short-term gains may be modest but over the next two to three years, the best is on a new kind of agency model — one that’s more responsive, more intelligent and, invariably, more automated.

Getting there, though, won’t be easy. Especially when it comes to the question that makes every holdco CEO flinch: remuneration. For businesses built on billable hours, technology that delivers more in less time is just as much a productivity story as it is an existential crisis. And that reckoning is already underway, quietly behind the scenes. 

“We’re not caught in time — as in we’re not incapable of changing our compensation models,” said Wren. “Our compensation models will increasingly shift, I think, to outcomes — however defined.” 

That shift is more micro than macro for now, playing out in select marketers under specific scopes with a handful of clients willing to experiment.