These are a list of FAQs to support the recent updates on Agentic Advertising and IAB Australia’s local sellers pilot. To read this article please CLICK HERE
Why should we care, and what is this actually trying to solve?
Because absence is not the same as ranking badly. Buyers already put software between a brief and a shortlist, and inventory a machine cannot read is not disadvantaged in that process. It is missing. Direct deals still move through email, PDFs and re-keying. That sets a floor below which a deal can cost more to service than it is worth, and much mid- and long-tail Australian supply sits under it. Agentic advertising is trying to solve a narrow problem: lower that floor, and give a buying system a way to evaluate inventory it did not already know about, without taking pricing or the spend decision off the seller. None of this depends on a forecast for how fast agent-mediated buying grows. The readiness work is the part we would defend either way: a media kit that is not only a PDF, floors that are not folklore, approval paths that are named. That is the on-ramp if a protocol prevails, and still an asset if none does.
Is the technology production-ready?
The reference implementations used in this pilot are a lab and sandbox surface, not the basis for a participant production rollout. That is why the programme is a sandbox: no live spend, no production writes, pinned software, human approval still on the path. The important work worth doing now is the catalogue, the floors and the gates – not a production cutover. The cheap time to be wrong about this is now.
What is AAMP?
AAMP, the Agentic Advertising Management Protocols, is IAB Tech Lab’s open-source framework for letting AI buying and selling agents work together using standards the industry already runs on, including OpenRTB, OpenDirect, AdCOM and the Deals API. AAMP 3.0 adds a new specification called OpenProposal, which standardises how sellers represent their ad products and respond to buyer briefs.
What’s the difference between AAMP 2.3 and AAMP 3.0?
AAMP 2.3, published in July, is the version with working reference implementations available today. AAMP 3.0, announced this week, adds the new OpenProposal specification, standardising how sellers represent their products and respond to buyer briefs. AAMP 3.0 builds upon AAMP 2.3 and doesn’t replace it. Both versions exist alongside each other, and AAMP 3.0 is currently in public comment until 22 October 2026.
One technical point is that the protocol and the reference software are numbered separately. AAMP 2.3 is the protocol release. The Buyer and Seller Agent SDKs that implement it carry their own version numbers, and the Australian pilot pins to seller-agent v2.4.2. Both numbers are current and they describe different things.
What does the AAMP protocol actually change day to day?
It gives buying agents a machine-readable way to discover and compare ad products across a wider range of publishers than a human planner could manage, and gives sellers a standard way to represent inventory, pricing and package rules so buying agents can read them accurately. It sits on top of existing standards and systems, it does not replace a DSP, SSP or ad server, and does not require publishers to rebuild their tech stack.
Does this mean AI is setting prices or doing deals without human oversight?
No. Commercial teams still set floor prices, package rules, discount bands and walk-away points. Agents operate inside those rules, and human approval stays part of the process at the points that matter, including agreeing a proposal and approving a price exception.
In the reference implementation the pilot runs, the approval gate fires on the proposal decision and on price exceptions. Booking-stage gating is a known gap in that implementation, which we record as a finding and score no participant against. Surfacing exactly that kind of thing is what a sandbox is for.
Could this replace publisher sales teams?
No. What agents take on is the repetitive coordination (checking availability, assembling a proposal that fits a brief, chasing the follow-ups) not the commercial judgement. Floor prices, package rules, discount bands and walk-away points are set by people, and human approval stays on the path to spend. The likely shift is that commercial teams spend less time on data entry and document assembly, and more time reviewing what an agent has produced and on the relationships and negotiations that need a person.
What is IAB Australia doing locally?
IAB Australia has been exercising the AAMP reference software in its own lab since early August and is opening a sandbox pilot for Australian publishers and sellers, with buyers, agencies and tech vendors invited to take part as observers, from 6 October 2026. The pilot will test discovery, proposal and approval workflows using synthetic data, with no live ad spend involved, and will feed local findings back into the global standard through IAB Tech Lab’s Agentic Task Force.
Why is IAB Australia testing AAMP 2.3 rather than starting straight on AAMP 3.0?
The pilot is pinned to AAMP 2.3 because it’s the version with working reference implementations, and IAB Australia has been testing it in its own lab since August. The first phase of the pilot, covering inventory modelling, pricing and approval policy and readiness assessment, isn’t protocol specific, so that work transfers regardless of which version comes next. There is no need to wait until late October; we want to be set up and generating learnings well before the end of the year.
AAMP 3.0 isn’t a requirement of this pilot. As AAMP 3.0 and OpenProposal progress, IAB Australia will separately run a controlled comparative workstream using the same synthetic data, commercial rules and core scenarios wherever they carry across. That work runs in IAB Australia’s own lab, not in participants’ environments – a participant’s pinned version doesn’t move unless the cohort agrees at a checkpoint.
Have you tested AAMP 3.0 or OpenProposal yourselves yet?
No. The OpenProposal specification entered public comment on 22 September and, per IAB Tech Lab, will be built into the Buyer and Seller Agent SDK reference implementations. There is no runnable AAMP 3.0 agent software to test yet. Our lab work to date has been against the AAMP 2.3 reference implementations, and nothing in the pilot has been tested against OpenProposal.
We have pinned the published repository, reproduced it in three environments and run the checks the repository ships with, and read the draft so that we can comment during the window. Reading a specification is not evaluating an implementation, and we will not describe it as one until there is code to run. In terms of design, it will give sellers genuine control over how products and responses are represented, and we’d like to be among the first to test it properly once there’s code to run.
What happens if AAMP 3.0 changes things mid-pilot?
Nothing changes in a participant’s environment without the cohort agreeing to it. The pilot pins deliberately and adopts upstream changes only at cohort-agreed checkpoints, so a release landing mid-programme can’t break a running scenario. Our comparative workstream runs alongside, in our own lab, and reports what a newer release changes rather than imposing it. If OpenProposal turns out to move the seller-side surfaces we hold evidence on, that is itself a finding and we’ll say so.
What have you found in the lab so far?
The findings report at the end of the pilot is the output, and we would rather publish conclusions than a running commentary on early tests.
What we can say now is that we’ve filed feedback consistently upstream with IAB Tech Lab both in GitHub and directly. One caveat we apply to our own results: the lab work so far has exercised deterministic workflow paths. It has not been a comprehensive evaluation of model-led reasoning or agent judgement, and we don’t claim that it is. That’s what the pilot is for.
Is real campaign data or real money involved in the pilot?
No. The pilot is sandbox only: synthetic data, no live spend, no production writes and no participation fee. The only real cost to a full-track participant is an LLM API key for their own agent, which we suggest capping at around A$100–200 for the programme.
How much time might this take from my team?
For the full track: an executive sponsor at around an hour a week, commercial and ad-operations coverage at two to three hours per working week if one person covers both (two to three hours each if the roles are split), and a technical lead at half a day to a day per working week during the hands-on phases. Readiness-first needs the sponsor and the commercial and ad-operations coverage, but no technical lead. Observers commit nothing beyond the time they choose to spend. Those figures are per *working* week, and none of them fall in the summer pause.
What do we get at the end?
Full track and readiness-first participants get a benchmarked readiness assessment against a six-level maturity model, and a structured inventory framework and machine-readable media-kit blueprint that is yours to keep. Full track adds hands-on sandbox transaction experience under your own guardrails; readiness-first sees the same flow in facilitated demonstrations. All three tracks get the shared cohort assets (the findings report, the Australian Seller Playbook and a member showcase) and a route into the global standard through IAB Tech Lab’s Agentic Task Force.
What’s in it for publishers commercially?
Publishers keep full control of what inventory is visible, to whom, and on what terms and a public rate card isn’t required. What the pilot tests is whether their inventory can be made machine-readable enough for a buying agent to find it. The commercial stake is that inventory a buying agent can’t read risks being absent from consideration altogether, not just ranked lower.
What does it mean for buyers, media agencies & tech vendors?
Buying agents can evaluate a broader set of publisher opportunities against a brief than a manual RFP process typically allows, and a proposal can carry the reasoning behind it (audience fit, format, terms) not just a price. Buyers, agencies, DSPs, verification and measurement vendors can join the Australian pilot as observers without deploying anything themselves.
Why does this matter for the Australian market specifically?
Australian internet advertising was worth $19.8 billion in FY26, up 14.0% on FY25 and its strongest financial-year growth since FY22, according to IAB Australia’s own revenue reports. As more buying activity moves through software reading product catalogues, publishers whose inventory isn’t machine-readable risk being left out of that demand.
Who can join the pilot, and how?
IAB Australia members can register interest by emailing agentic@iabaustralia.com.au before 5pm AEST on Wednesday 30 September 2026, nominating full track (standing up a seller agent, capped at six organisations), readiness-first (assessment and readiness work without deploying an agent), or observer (buyers, agencies, DSPs, verification and measurement vendors). Review runs 1 to 2 October, with places confirmed by 2 October, and kick-off is 6 October 2026. The programme runs for about 14 working weeks across roughly 22 calendar weeks, with a break over the Australian summer.
How many publishers can take part in the pilot?
The full track, where publishers stand up their own seller agent, is capped at six organisations, with one place reserved for a smaller or regional seller. Readiness-first participants, who complete the assessment and readiness work without deploying an agent, are counted separately and aren’t capped at a fixed number. The observer track, open to buyers, agencies, DSPs, verification and measurement vendors, is uncapped. The cap applies to the full track only, and it is set at six because the support offered is hands-on and six is the number we can do properly.
Which Australian publishers are taking part?
We are not naming participants. Expressions of interest close at 5pm AEST on Wednesday 30 September and places are confirmed by 2 October, so the cohort does not exist yet. Beyond that, participants are competitors sharing operational detail in a sandbox, and we do not identify them or confirm who has applied. If a participant chooses to talk publicly about its own involvement, that is theirs to do.
How does AAMP compare to AdCP?
Two open agent languages cover much the same jobs – discover, quote, book, report. Having tested both we see that they can run over the same agent-communication plumbing, but a shared transport does not reconcile advertising objects, permissions, commercial semantics or audit models, so they are best understood as different design philosophies rather than two layers that fit together. AAMP is the standards-body-led path built on the IAB standards Australian ad operations already run on – OpenDirect, the Deals API, OpenRTB – which is why our pilot is pinned to it. IAB Tech Lab’s Agent Registry is open to agents from any framework, and our lab has exercised both protocols on the same evidence discipline and filed feedback against both. The pilot’s core work, structured and machine-readable inventory with named floors and approval paths, pays off whichever language a buying agent arrives speaking. Our interest is whether Australian supply can be found and understood by whatever is doing the finding, not in picking a winner.
Who should journalists contact for more info?
Jonas Jaanimagi, Technology Lead, IAB Australia (jonas@iabaustralia.com.au)
