AI Has Started Buying Ads for You — IAB Just Set Two New Rulebooks

A few days ago, a friend of mine who works in ad buying invited me to dinner.

All through the meal, he kept sighing.

He said he'd just received a media brief: email a dozen or so publishers, wait for their proposals, then flip through them page by page comparing prices, schedules, and formats — negotiating back and forth, revising back and forth.

"Three weeks. Minimum."

I asked him: has this process been running the same way for thirty years?

He paused for a second, then said: yes, it really has.

Well, in the past couple of days, something fairly big happened in the advertising industry, and I have to tell you about it.

What Is AAMP?

Just in the past couple of days, IAB Tech Lab released AAMP 3.0.

What is AAMP? Agentic Advertising Management Protocols — the protocol suite for AI agents to manage advertising.

In plain English: it sets a common set of rules for the whole business of "AI buying ads on behalf of humans."

This 3.0 release added something new, called OpenProposal.

What does it do?

Think through the entire process of a media buy: the advertiser states its requirements, the publisher proposes a plan, the two sides negotiate terms, and in the end, commitments get made. That first stretch — from requirements to commitment — is, today, still mostly human work.

Emails. Spreadsheets. Meetings. More emails.

OpenProposal aims to standardize exactly this stretch.

It gives the seller — the publisher side — a unified, machine-readable format for describing its ad products and responding to a brief.

With this format in place, the buyer's AI agent can do the work on its own: discover what ad packages exist in the market, compare each one against the brief, check availability, evaluate fit, negotiate terms — and once a deal is struck, connect it to the existing execution and reporting standards, all in one flow.

What Does This Mean?

It means a lot more suppliers can now "be seen."

Think about it: how many proposals can a media planner seriously read in a day? Ten, twenty — that's the ceiling.

But an AI agent? It can pull every proposal that responded to the brief and run through all of them.

However small a publisher is, as long as it responds with a proposal in the standard format, it has a shot at making the media plan.

Sellers don't lose control, either. How their inventory is presented and sold is still entirely their call.

Of course, the rules build in guardrails too: pricing, orders, the identity of AI agents, repeat transactions — all of it is constrained.

Seriously impressive.

There's one more thing I need to set apart and talk about on its own.

AAMP 3.0 is not a rip-and-replace. It's built on top of IAB Tech Lab's existing stack of standards — AdCOM, OpenDirect, Deals API are all old foundations. The programmatic infrastructure already in place keeps right on working.

OpenProposal is now open for public comment, through October 22.

But IAB Did More Than One Thing This Time

At the same time, IAB also released another document: Measurement Services Addendum v1.0 — the measurement services addendum.

What is measurement?

An ad has run — do the impressions count? Who gets credit for the conversions? Can the data be trusted? All of this is what the industry calls measurement, spanning four service categories: measurement, verification, attribution, and analytics.

What's the chronic problem in this space?

Every measurement contract has to be negotiated from scratch.

Advertisers, agencies, publishers, platforms, measurement providers — every pairing negotiates terms with every other, an endless slog.

IAB's addendum gives everyone a common starting point: terms you can use as is — sign them directly if you're willing; if not, you can embed them into existing contracts and then adapt them to your own technology, methodology, and business model.

Together with IAB's earlier General Terms v1.1 and Direct Buy Addendum v1.0, it forms one "Lego-style" legal framework. From now on, contracts in the ad industry can snap together like building blocks.

Same drill: open for public comment, through October 22.

Finally, Back to My Friend

I told him: those three weeks of yours might really get shorter from now on.

No more emails to send, no more spreadsheets to compare. What you'll be doing is something else entirely: setting standards, building guardrails, and training your AI agent.

Standards don't create value directly; what standards eliminate is friction.

And the friction this industry has piled up over decades is now being dismantled, piece by piece.

Here's hoping your proposal soon stops sitting unread in someone's inbox.