There is one number in agentic advertising. Everything else is a roadmap.
Andrew Mole gave it. He is chief executive of pubX, a founding member of the Agentic Advertising Organization, and one of a very small number of people in this category running live agent-bought, agent-sold volume rather than describing it. In August 2026, he told ADOTAT the daily gross media spend was $2 to $3 thousand dollars. He answered in writing, on the record.
TL;DR. Andrew Mole of pubX is moving roughly $3,000 a day in live agentic gross media spend and is the only operator in the category willing to name an integer. He has also said, on the record, that most category announcements are the old programmatic infrastructure with a new word painted on them, and that this is harmful to buyers trying to evaluate real capability. The gap between his $3,000 and the trillion-dollar transition the category keeps predicting is not a contradiction. It is where the market actually is.
Why is $3,000 a day a useful number?
For context, $3,000 in daily gross media spend is roughly what a mid-market brand spends on search before lunch. The volume is not continuous; there was a break between campaigns, with the level returning the week the interview ran. It covers a handful of publishers across several sites.
Mole is not embarrassed by the number, and he should not be. He is the only person in this category who answered the scale question with a figure rather than a release schedule. Every other operator at this stage of the market defaults to describing a roadmap. Mole described a billing cycle.
He also went further than the number. Asked whether he stood behind his earlier account of companies putting $50 through a test, issuing a press release before the campaign had finished, and then reissuing the same press release six months later, he wrote: "yes I do stand by that and I think it's harmful."
The harm is specific. A term that should mark a genuine structural shift is being used to label incremental automation. When the label covers both "an AI assistant that helps a planner pull a brief together" and "two autonomous agents closing a deal with no human in the loop," the term stops carrying information. Buyers cannot ask useful questions. The few operators running real volume get measured against the same standard as everyone running a demo.
What separates agentic from programmatic with a new interface?
Mole made a second point that the category has been slow to absorb. A number of companies, he said, are laying an agentic layer over the infrastructure they already own. This is, in his assessment, a different thing from agent-to-agent buying, and it delivers no economic benefit to publishers or advertisers because it is literally the same programmatic system.
This is the diagnostic question that cuts through most agentic pitches: is the agent making decisions the old system could not make, or is it executing decisions the old system was already making, with a conversational interface attached?
Programmatic already had autonomy of action. A bidding engine sets prices millions of times a day with no approval step required, and has done so for well over a decade. The shift the agentic framing points at is different: autonomy of agreement, two agents closing a deal on terms neither side's humans confirmed in advance. That is genuinely new. Wrapping a campaign management interface in a chat layer is not.
The WPP Research simulation that modelled 90,202 agent-to-agent transactions found that the structural problem is not in the closing of deals, but in the recording of them. Under separate record-keeping, the two sides disagreed on at least one deal term in 95.3% of cases. Under a shared record both agents wrote to, that fell to 0.19%. The distinction between "an agent closed a deal" and "there is a shared, auditable record of what was agreed" is the same one Mole is pointing at when he says most category announcements describe the old pipes with a new word.
What does the Q1 2027 inflection actually mean?
Mole put the market inflection at Q1 2027, and his reasoning is structural rather than aspirational. Annual spend commitments locked into existing programmatic infrastructure elapse at the end of the year. That creates liquidity headroom for alternatives. The inflection point is a procurement cycle, not a technology readiness date.
Brian O'Kelley of the Ad Context Protocol, answering separately for the same ADOTAT investigation, described the current state of his own protocol with the same phrase Mole used: "not at scale." Shailley Singh of IAB Tech Lab, also on the record, said a model's word is not a transaction and that the architecture needs a deterministic audit trail that both sides can reference.
What does due diligence on an agentic claim actually require?
The test is simple to state. A vendor describing an agentic media buying capability should be able to answer three questions. First: what is the daily or weekly gross media spend running through the agent-to-agent layer, distinct from total platform volume? Second: is there a shared, machine-readable record of each transaction that both sides can query and receive the same answer? Third: has that reconciliation held up across a full billing cycle?
Most vendor answers will stop at the first question, because the number is either very small or not tracked separately from everything else the platform handles. Mole's $3,000 a day is the exception. It is small. It is also the most valuable number in the category right now, because it is real.
The category will scale. The procurement cycle will open. The operators with a verified record of what their agents actually agreed to will be in a different position from the ones who were running narration and calling it a transaction. The difference between those two things is what Mole is pointing at when he says the harm is real. The time to build the right infrastructure is before the scale arrives.