There is a phrase that occasionally surfaces when technology executives want to describe an industry gap between aspiration and execution. It is not a phrase one expects to hear at a trade conference, which may be precisely why Daniel Hulme, WPP's chief artificial intelligence officer, chose to deploy it.
Agentic AI, Dr. Hulme told attendees at the IAB U.K.'s AI Growth Summit, is in "the teenage sex phase" — everyone believes everyone else is doing it, but closer inspection reveals otherwise.
The observation arrives after a year in which the major advertising holding companies have announced AI operating systems, ad technology vendors have introduced tools for autonomous campaign creation, and chief marketing officers have delivered keynote addresses about futures that sound considerably more autonomous than present operations suggest. When the people who would actually deploy such technology — the traders, planners and campaign managers — are asked directly, the answers tend toward "exploring," "piloting" and "building the business case."
This is not, of course, advertising's first encounter with the gap between announcement and implementation. Programmatic buying was going to revolutionize media purchasing, and eventually it did, roughly a decade after the initial press releases. Data clean rooms were positioned as solutions to addressability problems that remain, for many practitioners, unsolved. (The pattern, by now, should be familiar.)
What distinguishes the current moment, Dr. Hulme suggested, is not that companies are misrepresenting their progress. It is that almost no one is being forthright about what deployment at scale actually requires.
"At least 80 percent of the energy that you need to build agents is testing," he said, adding, with a directness uncommon at industry gatherings, that deploying untested agents across an organization "is going to be a shit show."
The particular challenge for marketing, he noted, is that an agent trained on historical campaign data begins changing the very behavior it was built to predict the moment it starts running. Consumers respond differently, competitors adjust, media prices shift. The model risks obsolescence upon activation.
"You cannot predict the future based on the past," Dr. Hulme said.
None of which means the technology is not coming. It means, rather, that the industry is working with what Dr. Hulme characterized as the weakest version of AI — computers doing what humans already can, only faster. The genuinely transformative version, he suggested, involves systems that make decisions, observe outcomes and adapt.
By that standard, the advertising industry is not yet doing AI in any meaningful sense. It is doing very sophisticated rule-following.
Original story published in Digiday: "Why WPP's AI boss believes agents are still in the 'teenage sex' stage of development"