Agencies Think They Shouldn’t Own Agentic AI Tools Or The Underlying Data

For years, ad agency holdcos have been competing to build the best data and tech stacks.

Fast-forward to today, and agencies are looking to differentiate their tech stacks by building custom agentic AI tools for their brand clients. And they seem less interested in owning the data that powers the AI. They’re even rethinking owning the AI tools themselves – particularly since licensing them to clients creates new revenue streams.

A lineup of agency heavy hitters discussed the changing buy-side landscape at Programmatic IO in New York City on Monday. The panel included Sir Martin Sorrell, founder and executive chairman of S4 Capital, which owns Monks; Bob Lord, president of Horizon Media; Maggie Summers, chief experience officer at dentsu; and Shane McAndrew, chief data and technology officer at WPP Media.

Who owns the data?

Agencies owning data creates more headaches than it’s worth, the panelists largely agreed.

“Ownership of data brings problems,” Sorrell said, because it opens agencies up to accusations that they’re grading their own homework.

Meanwhile, advertisers are always wary of black boxes when it comes to their campaign optimization. And with dentsu expecting algorithms to drive 75% of ad spend by 2028, Sorrell said, it’s put more pressure on agencies to prioritize transparency and neutrality in how they work with data partners.

This shift seems to buck the recent trend of agency holdcos buying up data and technology platforms. Publicis’ technology strategy – including its 2019 acquisition of Epsilon to unify client data across its business – put it at the front of the holdco arms race at the end of 2024, according to investment analyst Terence Kawaja. But Publicis’ rivals alluded to its ad tech acquisitions as a liability on the Programmatic IO stage this year.

For instance, Sorrell didn’t pass up a chance to take a shot at Publicis’ planned acquisition of alternative ID and audience data provider LiveRamp and whether that relationship makes LiveRamp less of an objective source of truth. When AdExchanger Editorial Director Sarah Sluis asked Sorrel how he would advise a client to work with LiveRamp once Publicis’ acquisition of the company closes, he responded, “We’d recommend clients work with the best possible alternative.”

From left: AdExchanger’s Sarah Sluis, S4 Capital’s Sir Martin Sorrell, dentsu’s Maggie Summers, WPP Media’s Shane McAndrew and Horizon Media’s Bob Lord onstage at Programmatic IO NY 2026.

Horizon Media is similarly backing away from data ownership, according to Lord. He said, when it comes to the agency’s blu. AI platform, the brand client owns the underlying data and also owns any AI models created on top of that data.

“The client has to retain that data ownership, because then, if you ever have a breakup, the agency can’t take that data with them,” Lord said. He added that he’s seen brand and agency breakups throughout his career where the agency has taken the brand’s data and applied it for a competitor’s campaign – which, he added, is basically like taking a client’s ideas and handing them over to a rival company.

And, as agentic AI adoption grows, Lord said he’s also seen other agencies treat the AI tools built on client data as the agency’s IP that can be used for other campaigns. “That’s not acceptable,” he said.

What replaces the FTE model?

In addition to addressing how AI is changing data partnerships, the execs weighed in on how AI is transforming how agencies do business at more fundamental levels.

For example, AI is already changing how agencies work with DSPs, said WPP’s McAndrew.

“If you think about who is using DSPs, we’re starting to see it’s less human and more machine, which is great,” he said. “We can go agent-to-agent and get more intelligence than we’ve gotten out of those DSPs faster.”

But, he added, AI is also a threat to DSPs because it makes it easier for buyers to get strong data signals from SSPs and the publisher side, cutting out DSP data marketplaces in the process. “There’s always pressure to compress [via] supply-chain optimization and find efficiencies, and you’re going to see AI accelerate that.”

Agencies are under pressure from AI, too, Sorrell said. He pointed to Meta’s plans to create an end-to-end creative, planning, buying and distribution platform as an existential threat to the enterprise agency model. He predicted agencies will have to pivot to close partnerships with Big Tech’s end-to-end model, and then “the agency will increasingly become the validator for that model.”

In that world, though, the legacy agency FTE compensation model, where agencies charge based on how many people are assigned to a client account, makes less sense, Lord said.

However, agencies have to resist the urge to see AI and the automation of manual marketing tasks as an opportunity to offer client savings by cutting headcount, McAndrew said.

“We’re not going to cost cut our way to growth, and we’re squarely in the growth business for our clients,” he said. “We will identify the high-volume, low-value paths, and we’ll automate that – that will just get done. But 95% of our focus at WPP is on delivering the right set of intelligence to our clients to drive that growth, measure that growth and validate it.”

Instead of FTE-based pricing, Sorrell advocated for agencies to shift toward outcome-based pricing for campaigns.

For his part, Lord advised agencies to “create a P&L focused around the client, not the silo.” He said Horizon’s collaboration with Havas, called Horizon Global, focuses on closely serving large global clients.

There’s an opportunity for agencies to focus less on managed services and to prioritize “more hands on keyboard” – or, in other words, licensing software directly to clients, Lord said. He added that introducing new agentic interfaces allows agencies to “charge our clients for something very tangible that they’re logging into.”

But, Lord added, “the inertia in the system, the FTE-based model which the consultants put us on, hurts us” and prevents agencies from shifting to newer business models.

While change takes time, outcomes-based models will eventually win out, said dentsu’s Summers. Dentsu is currently operating a mix of the old FTE-based model while experimenting with new outcomes-based approaches, she said.

“It’s this hybrid of paying for the intelligence, for the people who are generating the ideas, building the experiences and orchestrating them,” she said, “and then have a model for outcomes so that we share and appreciate the value of that outcomes-driven model.”