Programmatic Pioneers 2025: Sorrell, industry leaders predict AI-led shake-up in media buying roles

Barely a day after soon-to-be former WPP CEO Mark Read said there was “no doubt” AI would “compress” many advertising jobs, his predecessor, Sir Martin Sorrell, echoed the warning, predicting that AI would “completely gut what goes on in media planning and buying.”

The two comments were made at different conferences — SXSW London and Programmatic Pioneers — but they raise the same critical question: What does the media buyer of the future look like?

Sorrell drew a parallel with financial markets, asking, “BlackRock or Blackstone … do they do things manually or semiautomatically?” He noted that out of the estimated $1 trillion in global ad spend, most of it is not traded algorithmically. That may soon change.

“We won’t have 25-year-old media planners deciding where hundreds of millions of ad dollars go,” Sorrell said.

So what’s the role of agencies in this future? “To work very closely with clients to validate what the algorithms say,” Sorrell said. “But in terms of employment, huge compression. It’s going to be half of that number [250,000 employees in agencies around the world]. Because it’s all going to be machine-driven.”

Dystopic vision or forthcoming reality?

Big Tech — led by Meta — is allegedly already working to make fully automated ad buying happen from creative execution to targeted delivery.

Still, when it comes to the creative side of the business, Sorrell speculates that an onslaught of AI-created ads may actually mean more employment for some agencies. Though in the short term, there may be “fewer copywriters and art directors.”

Algorithms may be able to buy and trade advertising, but “if everyone is producing personalized ads at scale, how do you differentiate?” That will put “a premium on creativity,” he said.

Timo Schulte, COO at Oscar Bravo, Lufthansa Group’s in-house agency, also mused that soon there “won’t be a job like a programmatic trader setting up a campaign.” Schulte imagines that role could instead be replaced by a “marketing prompt engineer.”

“In the last couple of years, we’ve focused on targeting and reporting data, but I believe that training data is the key to performance. We have to give the right training data to the right model, and the rest is just automation and performance. … There’s no campaign anymore. There’s just an engine,” Schulte said onstage.

Whose side is the machine on?

Marketers who plan to use fully AI-automated platforms may be implicitly choosing to accept a certain level of opacity.

“If you trust the machine, at a certain point, you also have to accept the outcome,” Schulte said.

But issues can arise when it’s not exactly clear whose side the machine is on. The advertiser’s, or the platform’s? What if the platform also owns the media where the ads are displayed?

In this respect, some marketers wondered why walled gardens’ fully automated AI-buying systems weren’t being held to the same transparency standards as premium media on the open internet.

Even though walled gardens ostensibly allow for third-party measurement, “If you’re allowed by Google to measure YouTube, are you going to say YouTube is bad? No, because you’ll be kicked out,” said Jacques Edeling, head of programmatic at Starcom.“There’s definitely a conflict of interest in many ways. Why is [a lack of transparency standards] OK in a walled garden but not outside the walled garden?”

AI may just not be quite there yet

Ricky Chopra, global head of innovation and AI at Zenith, said that AI simply isn’t there yet to be entrusted with the big budgets.

“Would I trust an AI with a million-dollar ad campaign? No, and I don’t expect our clients to. I can’t imagine anyone accepting the excuse, well, a machine made that decision, we’re not in control anymore.”

For now, Chopra thinks there will always be a human in the loop.

“We had to employ people who understood how to speak machine. Now the machines are understanding how to speak human, and so now what we need is people to speak human, people who understand what the challenge is, what they’re trying to achieve, and have a conversation with a machine.”