Social Rocket
The IAB Tech Lab has extended the public comment period for AAMP 3.0 and OpenProposal to 10 November 2026. The updated official announcement explicitly names the new date; the older body text still gives the original deadline of 22 October. For agencies, the question behind this is more important than the date: can AI agents compare advertising inventory more transparently in future without losing control and cost oversight over purchasing?
What OpenProposal is actually meant to standardise
The version AAMP 3.0, published on 22 September 2026, adds OpenProposal to the existing IAB standards. Providers should be able to describe their advertising packages in machine-readable form and respond to briefings. On the buyer side, agents can discover, compare and check offers against audience, duration, format and terms.
This is the phase before the actual purchase. Existing standards such as OpenDirect, AdCOM and the Deals API remain relevant for contracting and execution. The IAB is therefore not promising a new universal advertising network, nor the abolition of all DSPs. The technical overview from the IAB Tech Lab also provides reference SDKs for buyer and seller agents as well as an agent registry.
Why smaller agencies are looking closely at fees
The pressure to act is evident from a Digiday discussion from 27 October: independent US agencies complain about access and intermediary costs charged by established demand-side platforms, including The Trade Desk and Amazon DSP. These statements are experience reports, not an independently verified fee schedule applying to all clients.
Open offer formats can make it easier to search for and pre-screen new inventory sources. Whether this ultimately leads to lower commissions, cheaper CPMs or better campaign results is not proven. An AI agent can even add costs if platform fees, service-provider margins, data costs and model operation coexist.
Five checkpoints instead of autopilot
A sensible agency pilot should document the origin of every offer and report four quantities separately: net media price, platform and reseller fees, data/technology costs and actual performance metrics. Add a human approval step before binding booking, consent and brand-safety checks, and a traceable log of every price change. Without these rules, faster buying does not automatically mean better buying.
The IAB Tech Lab describes technical safeguards explicitly against double bookings on repeated agent calls. Precisely these unglamorous control rules are decisive for productive use.
Our report on CNN's planned ad sales via AI agents also shows that real deployments are only just getting under way. Which role classic buying platforms retain is illustrated by Pinterest's programmatic integration. On auction mechanisms, it is also worth looking at Prebid's proposal for the ad server market.
Classification: AAMP 3.0 is a standardisation proposal under ongoing consultation, not evidence that fully autonomous media buying works industry-wide today.