Ad Tech Briefing: AI is changing programmatic buying faster than who controls it

In ⁠its 2026 study, IAB Europe found that 86% of 50 respondents use AI for marketing. Among the 29 asked about specific advertising workflows, 59% selected programmatic optimization and 55% media planning and strategy. Agentic buying and selling ranked lower at 38%. The distinction matters: programmatic has long relied on algorithms, while agents that plan and execute transactions with varying degrees of human approval remain less established.

Expectations are ahead of deployment. Asked what would happen over the next year, 30% expected agents to become a main method of buying and selling ads in some markets, while 28% anticipated regular use without that level of scale. Another 18% expected further development but little day-to-day use. No forecast commanded a majority.

Human oversight

Of 47 respondents answering a question about autonomy, 36 reported either no agentic system in everyday use or one at the study’s first two levels, where people direct the agent or plan alongside it. Larger organizations reported more advanced systems on average than smaller ones, while ad tech respondents reported greater autonomy than agencies and advertisers. The small groups behind those comparisons make them signals to investigate, rather than a ranking of the industry.

The evidence for performance is similarly mixed. Respondents who rated AI in current ad operations gave it an average of 2.76 out of five. Among 29 respondents asked how they evaluate AI tools, 69% selected operational efficiency and 48% CPM improvement; just 21% selected better audience quality or targeting accuracy, and 3% fill rate or yield. Five said they had no formal measurement. Those answers suggest that saving time and lowering media costs are easier to demonstrate than a sustained improvement in what an auction earns.

The buying stack is also unlikely to become purely proprietary or purely platform-run. Respondents described a mix of in-house and third-party tools, and of AI built into ad platforms and supplied separately.

Of course, this raises the critical issue of whose system makes a buying decision in any buying campaign, along with who can inspect it, and who takes responsibility when it affects spend. Although 78% of the full sample identified an AI governance owner, only 48% had marketing-specific AI guidelines. Among organizations providing training, fewer than a quarter covered checking autonomous actions and knowing when to intervene.

Google’s updates

Those unresolved questions about control extend to publishers on the other side of the auction, with a recent ⁠LinkedIn post from MonetizeMore sparking a debate over a newly documented Google Ad Manager option to set price floors for specific programmatic bidders.

The post frames that option as the return of a lever publishers lost when Google introduced unified pricing rules in 2019. The court’s ad tech remedies opinion describes those rules as having removed publishers’ ability to set a higher floor for AdX than for rival exchanges. Google’s ⁠current instructions show how to select individual bidders in a pricing rule, though the parties are still working out the language of the final judgment.

Subsequent debate from the original post centers on whether publishers can use different floors to put more pressure on selected buyers or reward demand sources they value. Its practical warning is to calculate against net publisher revenue: Google’s own example says a $1 bid under an 80/20 revenue split is tested as $0.80 after its share. The post also asserts that bidder-specific floors remain hidden from buyers; Google’s documentation explains limits on the floor disclosed in bid requests, but does not explicitly confirm every part of that bidder-specific claim.

The question it poses fits the survey’s wider finding: as more decisions become automated, the value of a control depends on whether publishers can see its effects and use it without sacrificing worthwhile bids.

What we heard

“Meta has spent years learning what its users are interested in and how to create demand. Now, by taking aim at the checkout experience, Meta is attempting to collapse the entire funnel into one interface.”

– Viant CEO Tim Vanderhok on “the most consequential development in ads + agentic commerce,” via LinkedIn.

Numbers to know

Further findings from the IAB Europe study:

• 76%: Of 29 respondents asked about advertising workflows, 22 use AI for reporting, analysis or dashboards.
• 42%: Of the full 50-person sample, 21 cited a lack of expertise or training as a barrier to AI adoption.
• 24%: Of 38 respondents whose organizations provide AI training, nine said it covers checking autonomous actions and knowing when to intervene.
• 62%: Of the full sample, 31 said industry guidance on AI for marketing would be useful.

What we covered

‘Not hiring’ entry-level talent: Industry acknowledges runaway crisis brought on by introducing AI to the workforce

As many inconsistencies exist across adland that AI has introduced, executives seem to agree that the pace of AI adoption has created a talent crisis. 

Marketers bemoan the culture of ‘good enough’ as standards fall amid the AI bubble

Marketers and their advertising teams are seeking to implement new standards and build a talent base fit for the challenges of the contemporary era.

What we’re reading

Amazon blocks Meta’s Muse from shopping on its site

Amazon says the agent accessed its store without authorization, sharpening the question of who controls agent-led commerce.

⁠McDonald’s joins the advertising business

The chain is testing ads on drive-thru menu boards and exploring how to turn its customer data into a media business.

Citi is jumping into the advertising business

Its new commerce media unit will let brands target customers using spending data, adding another financial institution to the retail media field.

Scope3 rebrands as Apostra and pivots toward agentic ad buying

Cofounders Brian O’Kelley and Anne Coghlan are repositioning the company around advertising transactions involving AI agents. (⁠businessinsider.com)