‘Humans at the helm, agents in the loop’: WPP is testing a buyer ageny for video, but it’s bigger bet is on governance

“Autmation for us isn’t the goal, better decision making is the goal,” said Lauren Wetzel, global president of data and technology solutions at WPP.

That promise is partly technical and partly historical. Every media owner in the coalition is building its own seller agent on its own infrastructure, using its own models. Without a common standard between them, the risk is a new layer of fragmentation that’s more opaque and harder to audit than what came before. Wetzel called it the “Rosetta Stone” problem, So alongside building the agency, WPP Media is working with its partners and standards bodies to establish shared protocols for how buyer and seller agents communicate, validate decisions and execute approved transitions. That work, she said, is as important as the agent itself. 

Unsurprisingly, the specifics are still being worked out. But at a broad level the standards cover three areas: the underlying technical communication between agents, the mechanics of media buying itself and the particular complexities of video and CTV transitions. Audit trails and approval thresholds run through all of it. WPP holds a board seat on the IAB Tech Lab and is an active participant in Prebid.org, giving it a direct feedback loop into the rules being written. 

The agent itself operates within that structure, evaluating inventory opportunities, applying audience and performance intelligence as well as making planning recommendations. From there, proposal and approval workflows can extend to programmatic guaranteed and private marketplaces through to activation thresholds. Financial commitments and campaign activation require explicitly human sign off. That way, the decisions stay with planners and buyers — for now. 

Wetzel was candid that the thresholds will shift over time, based on where testing shows agents outperforming humans at each point in the workflow. According to the company, findings from testing will not be published for broader industry use until early 2027.

“Over time we’re going to make changes on what the role of governance is based on what we see performing better at what point in the workflow,” she said. “That’s a really important part of the testing as well — who’s the better decision maker on this. In some instances it’ll probably be the humans, especially at the starting point.”

It is an honest admission about what agentic media buying ultimately means for agencies. Not the removal of human expertise but a slow negotiation over where it still has the edge. That doesn’t mean they’re a fail-safe. Rather, she’s describing people as the better decision maker, at least for now, with the balance of judgement between human and machine to be worked out through testing rather than assumed in advance. 

“It’s accountable media buying, humans are at the helm, agents are in the loop,” said Wetzel.

In its current form the agency operates at scale across multiple advertisers rather than being configured individually for each one. The client-specific version, where marketers can direct more of the decisions themselves, comes later. 

“You’ll see as this technology evolves that there may be some nuances to some of the challenges that our clients have,” said a WPP spokesperson.”When we have clients that are particularly interested in having more ability to direct some of those decisions, then that is something that is part of where we could go.”

But the standards and governance only matter if the intelligence feeding the agent is worth acting on. After all, the industry has historically relied on static, off-the-shelf data that is stale the moment it hits a campaign and available to every competitor equally. An agent running on that kind of data is not making better decisions. It is making the same ones, faster. Wetzel’s argument is that Open Intelligence — WPP’s large marketing model and data solution — changes that calculus, pulling in faster more dynamic signals, past performance data, social listening and contextual signals that legacy identity infrastructure previously could not match. 

The goal is what she called an “asymmetric advantage.” What the agent does for one brand should be materially different from what it does for another. Each client’s data remains siloed and inaccessible to other advertisers, though the full degree of per-brand customization is a next stage ambition. 

“Better decision making only comes from the intelligence and the signals that you’re able to feed and contribute into giving direction between buyer and seller agent,” said Wetzel.

The restraint is deliberate. Wetzel points to programmatic as the cautionary tale and does not exempt WPP from it. The industry introduced the technology, the challenges followed and then spent years trying to fix what it had broken. The current discourse around agentic media buying risks repeating the same pattern. Too many incentives, rivalries and commercial interests need to align for any structural shift to deliver cleanly on its promise. The smarter publishers already know this. News U.K. and CNN are among those developing their own seller agents, ensuring they can meet buyers on their own terms rather than waiting for the industry to bring buyers to them. 

“I wouldn’t ever claim that agentic media is going to save publishers, and I also don’t think that it’s going to kill publishers,” Wetzel said. 

What the agent can do is create more efficient and intelligent ways for buyers and sellers to interact, potentially surfacing new supply and ensuring premium inventory is represented at its true value in an increasingly automated marketplace. The structural challenges facing publishers are, in other words, not WPP Media’s to solve.

“I call our publisher partners our intelligence partners,” Wetzel said. “I don’t think of them as simply inventory partners.”

They will not have to wait long to find out how seriously WPP means it. 

Testing between WPP Media and its partners is already underway, though at an early stage. Rather than anchoring to a single partner or standard, WPP has been testing across multiple platforms simultaneously, working out workflow refinements as it goes. The near-term target is large-scale TV and video buying capability within six to nine months, with handling bigger client budgets the immediate proof point.

Video and CTV are the starting point precisely because they represent the hardest version of the problem: the most complex governance, the largest transactions, the biggest budgets. Get that right, the thinking goes, and the rest becomes easier. 

“As both sides of the ecosystem become more confident, our testing can scale alongside that,” said the WPP spokesperson “Each market has its own providers, its own nuances that we’ll have to navigate and scale along with our clients.”

The long-term ambition extends well beyond video. How far beyond, WPP is not saying. 

What Wetzel will say is that she does not see other agencies building. She sees them attending the meetings, participating in the working groups, sending representatives to the bodies writing the rules. But the actual construction, the engineer-to-engineer work of making buyer and seller agents speak to each other, she says that is happening at WPP and not many other places. “I don’t see other agencies taking those steps,” Wetzel said. “I see a lot of participation in the industry standards, but I don’t see the innovation and the development happening.”

Two decades ago, the companies that moved first in programmatic built businesses that defined the industry for a generation. Some of those businesses also caused significant harm. The question the industry is now trying to answer is whether it is possible to be first and careful at the same time. WPP Media would like to believe the answer is yes.

“Wpp open will be attractive to many advertisers,” said Robert Webster, founder of AI marketing consultancy TAU and a former WPP exec. Though many will not want to rent out their marketing brain and prefer control and transparency. My big concern in this approach is WPP has many inhouse owned media options where they take margin. How can you trust an agency AI which is tied to margin? The google DOJ case showed us this is a bad idea.”