The case for and against agentic media buying

And while that may sound good to publishers, open-exchange programmatic advertising has also taught them not to trust the upside.

The Trade Desk’s Jeff Green said in an earnings call on April 7 that agentic buying only works at the scale of the open marketplace, not through direct buyer-seller relationships — a model he suggested some competitors are mistakenly building toward. 

“Most companies that are focused on agentic in our space are just talking about plugging into these tiny pools of inventory,” Green told analysts on the call. “One advertiser connects to one publisher, and in doing so, you more or less create another ad network, where you have hundreds of thousands of ad networks because of the combination of one advertiser to one publisher and agents talking to each other.”

Holding companies have been steadily consolidating supply paths, leaning harder into curated PMPs and bringing more activation in-house through their own curation and supply-path optimization layers. The open exchange is getting hurt in the process. Whether agentic media trading can reverse that remains an open question. 

Here’s a look at both sides:

The case for agentic media buying 

Many publishers regard agentic buying as a structural fix for the bloated programmatic ad supply chain. By stripping out layers of ad tech toll-takers, agents could push more budget into actual working media and back to publishers, noted Paul Bannister, chief strategy officer at Raptive.

“Agentic media trading does stand to contribute to supply chain cleanup, as it puts the buyer directly in touch with the publisher’s ad server and removes other middlemen from the transaction,” said Justin Wohl, vp of strategy at Aditude and consultant for publisher Salon.

At scale, agents have the potential to ruthlessly reprioritize cheap, commoditized exchange inventory, according to Gabriel Dorosz, advertising lead at publisher association INMA. “Once buy-side agents start operating at scale, commoditized open exchange inventory becomes the easiest thing to optimize away from,” he said.

That means that agents could accelerate a flight to quality, but in the process, they’d drain oxygen from the rest of the open exchange. Good news for top-tier, data-rich publishers; bad news for the long-tail open web. “The premium inventory, the contextual and first-party-data-rich inventory, the inventory with verified attention and outcomes, will draw demand at the expense of everything else,” he said. 

Wohl agrees agentic trading could even bump up yields. “It could lead to a shift in total revenue coming from programmatic back to ‘direct’; that’s how I’d classify the agentic deals that book the publisher’s inventory directly. It might even be at a premium CPM to what programmatic transacts at,” he added. 

In theory, agents can automatically optimize around performance signals and strip out manual inefficiencies for buyers. If brands insist on transparency, they can get the same or better control they have manually, but at machine speed, said an ad tech exec, who exchanged anonymity for candor.

Meanwhile, properly governed agents can be trained to favor compliant, high-quality data and environments, not just the cheapest CPM, stressed the same exec. They can be instructed to chase incremental reach — new customers — rather than endlessly retargeting the same users, which supports more robust mixed-media modeling and proof of performance.

Because AI accelerates what’s already there, strong audit and procurement processes can use agentic buying to reward ethical, transparent partners and marginalize the murky ones faster. 

The case against agentic media buying

Some worry that agencies and intermediaries will use agents to reintroduce obfuscation, just as the ANA/PwC work has nudged programmatic toward clarity. Their reasoning is that if buying agents are simply optimizing to an outcome while talking to opaque selling agents, clients will lose sight of where ads actually run, what data is being used and whether it’s compliant — or whether you’re reaching incremental audiences versus hammering the same people.

Used badly, agentic buying becomes “outcomes at any cost,” said Craig Tuck, chief revenue officer of Ozone. This would just accelerate the old margin-grabbing, low-quality open web behaviors under a new label. 

“No doubt it will be used for a whole amount of get-rich-quick obfuscation [and] intermediary gains,” he said. “That’s just digital advertising with a new name — you’ve got practitioners who are compliant, ethical, do things properly, and a whole subsection that don’t care about that.” 

Some agency buyers argue that agents shouldn’t be used at all for autonomous buying, but be restricted to assisting media planning and campaign setup.

“Agentic buying tools could improve transparency around repetitive execution tasks, but there’s real risk they could also obscure critical decision-making if advertisers rely too much on AI for bidding, optimization, and performance assessment without human oversight,” said Amy Porter, svp and executive director of digital media at RPA.

“The future likely isn’t fully autonomous media buying, but instead it could be AI augmenting operational workflows while experienced practitioners remain responsible for strategy, judgment, and accountability,” she added.

While there have been, and continue to be, a flurry of agentic buying tests, the reality is that agentic trading is still tiny, operationally messy and far from proven. “It might be aspirational where we want to go to as an industry… I don’t think we’re there yet,” as the Daily Mail’s managing director of advertising Pierce Cook-Anderson put it. 

That’s pretty much the consensus. “It’s still super nascent — there’s very little money being transacted,” said Bannister. He said that Raptive currently has a proof of concept agentic trading tests set up, but only exchanging small budgets — “in the dollars,” he added.

“Changing people’s behavior, giving people new tools, giving people new ways to look at things, making sure the million things that are set up in ad tech are also set up in agentic — it’s all complicated, and it takes time,” said Bannister. 

For all the optimism about agents cleaning up the ad tech pipe, neither Bannister nor Wohl see them as a cure for the open web’s deeper malaise: a broken value exchange where platforms and AI tools increasingly keep the value and starve publishers of traffic. 

Bannister sees agents as almost like a sideshow to the collapse of the old content-for-traffic value exchange as platforms and AI systems scrape, synthesize and keep users within their own surfaces. “Previously, publishers gave their content away to platforms, and then [they] drove traffic back. Now, platforms steal the content, and they don’t send traffic out. That value exchange is broken — the agents, scrapers, are a symptom of a bigger problem,” he added. 

So, while agents might marginally improve how money moves through the pipe, they won’t on their own stop the open web’s underlying erosion. “Sadly I don’t inherently see this as a saving grace for publishers, not if they keep losing their audiences to generative AI alternatives to their websites,” said Wohl. “It can’t change the traffic trajectories.”

—Seb Joseph contributed to reporting.