Building the Agentic Open Web: IAB Tech Lab's Push for Shared Standards

Image from IAB Tech Lab Summit courtesy of IAB.

The conversation at the recent IAB Tech Lab Summit made one thing certain: agentic advertising is rapidly moving from a theoretical future to a practical engineering reality. The upcoming July release of AAMP 3.0 (Agentic Advertising Management Protocols), backed by a working reference implementation already running in the cloud, provides the industry with its first concrete blueprint for machine-to-machine transactions. As autonomous buyer and seller agents begin to coordinate campaigns directly, the traditional premium on platform-specific execution will start to diminish. For brands and agencies alike, long-term competitiveness will rely less on manual platform mastery and more on becoming highly structured, trusted partners that autonomous systems can seamlessly read, verify, and select.

The Challenge of Interoperability

If Google Marketing Live was a preview of what an agentic ecosystem looks like inside a single walled garden, the IAB Tech Lab Summit was a preview of what it has to look like everywhere else. Anthony Katsur, CEO of IAB Tech Lab, framed this evolution as “From Turing to the Agentic Web” a shift from predictive machine learning to fully autonomous agents that coordinate decisions and workflows across the ecosystem. The question the room kept circling back to is straightforward: how do you make that work across thousands of buyers, sellers, and intermediaries without devolving into chaos?

The answer is that joint standards are the only viable path to scale. Interoperability won’t be figured out through MCP (Model Context Protocol, how AI models securely connect to external tools, databases, and file systems) on its own. Neither will thousands of bespoke agent integrations. AAMP is IAB Tech Lab’s bet that an open, shared protocol layer is what unlocks the value of agentic AI for the open web. Here is our read on what was announced, what it means for brands and agencies, and where we’re watching closely.

Why Standards, and Why Now

Four themes came up in every session. Together, they make the case for why agentic advertising can’t scale without shared rails.

Accuracy. Agentic chains introduce semantic drift, a game of telephone where small misinterpretations between agents compound into a meaningful gap by the end. A brief that loses nuance across several agents can lead the system to purchase inventory that departs from the original intent.

Cost and efficiency. Tokens aren’t free. Poorly organized inputs create token bloat that inflates AI processing fees and widens the surface area for hallucinations. A shared protocol functions as a shorthand, so agents don’t renegotiate vocabulary every time they speak.

Transparency and auditability. Agentic workflows make a lot of decisions fast. If humans can’t reconstruct what an agent intended, what it considered, and what it did, we won’t be able to defend the work to clients, regulators, or platforms. In the wake of a renewed focus on supply chain transparency this year, auditability has real business implications.

Security. The same protocols that let agents transact also have to keep bad actors from impersonating them. Sandboxing, cryptographic verification, and explicit intent declarations sit in the guardrail layer beneath the stack.

AAMP: The Architecture for an Agentic Open Web

AAMP is the umbrella initiative that ties IAB Tech Lab’s agentic work together. Three pillars sit on top of each other: agentic Foundations at the base, agentic protocols in the middle, and trust and transparency at the top. Version 3.0 lands this July, and a few pieces are worth flagging.

Buyer and Seller Agent SDKs. Both sides of the transaction get a parallel three-tier hierarchy. A Portfolio Manager agent handles strategy. Channel Specialists translate that strategy into CTV, display, video, mobile, and native. Functional agents do the bounded tasks like research, audience planning, bid execution, pricing, and avails. The hierarchy matters because it separates the decisions an LLM should be making from the ones it absolutely shouldn’t.

Protocols built on what already works. AAMP is deliberately not a greenfield project. Its components extend existing IAB Tech Lab specs like AdCOM, OpenDirect, OpenRTB, and the Deals API, plus LiveRamp’s User Context Protocol for audience signals.

An Agent Registry for trust. The top of the stack is an industry-wide source of truth for verifying who or what is on the other side of a transaction. Pair that with a Programmatic Governance Council to keep the rules current, and the ecosystem gets the human readable backstop it needs.

“Just Build It”: Deploying the Standard in the Cloud

One of the most useful sessions at the summit was a live walkthrough of an open-source reference implementation, with a buyer agent and a seller agent talking to each other on top of Amazon Bedrock AgentCore.

In the “Agentic Arena” demo, an agency agent and an AAMP seller agent coordinated a campaign end to end. The systems automatically handled audience activation, deal acceptance, creative requirements, pacing alerts, and budget adjustments, while human operators only stepped in for strategic exceptions like brand safety incidents, IVT thresholds, and major pacing swings. It made the abstract feel concrete. In practice, the agentic campaign console functions as a chat interface for orchestrating a coordinated crew of specialized agents. The path from spec to production has gotten short, and the brands and agencies that start experimenting now will have informed opinions when 3.0 lands.

From Keeping Bots Out to Letting Agents In

Digital advertising spent two decades engineering bots out of the funnel. Bot traffic was fraud, full stop. The summit forced an intriguing reframe, because the agents shopping, comparing, and transacting on behalf of consumers aren’t fraud. They’re an extension of the human behind them. We have to let them in and let them play their role, while still distinguishing them from the bad actors that invalid traffic detection and bot management were built to block.

That reframe has a measurement consequence. Identity resolution can’t stop at cross-device anymore. It has to extend to cross-entity, reconciling a person and the constellation of agents acting on that person’s behalf as a single coherent self. Vector embeddings will do a lot of the quiet work here, representing identity, intent, and context as points in a coordinate system so agents can pass meaning to each other without re-litigating definitions on every hop. UCP (Universal Commerce Protocol), now formalized into AAMP’s Agentic Audiences pillar, is the first industry wide attempt to standardize that exchange.

What This Means for Brands and Agencies

Deep channel expertise is a depreciating asset. If a CTV agent or DSP agent is handling channel specific execution, the value of being the person who can navigate one platform’s UI drops fast. The new value lies in being the strategist who can manage a crew of agents across the bigger picture.

AI cares about data, structure, and credibility. There’s a widening gap between how brands talk about themselves in paid media and in the sources LLMs treat as authoritative. If your most structured, machine-readable content lives in a PDF on the eighth tab of your IR site, agents will quote your competitors’ press releases instead. Audit what’s credible to a machine, not just what’s on brand for a human.

Enrich the intention data agents inherit. Briefs used to be written for humans who could read between the lines. Agents can’t if they don’t have the right context to draw from. Investing now in structured, machine-ingestible briefs (ensuring clear objectives, audience definitions, creative constraints, guardrails) pays back the moment a Portfolio Manager agent is the one assembling your campaign.

Get a seat at the protocol table. AAMP is open source and developed through working groups. The brands and agencies contributing to the buyer and seller task forces, the Programmatic Governance Council, and the AAMP 3.0 review will shape how the standard evolves. Watching from the sidelines is a strategy with a known outcome.

The Road Ahead

Google Marketing Live made the case that agentic AI is coming to the campaign console. IAB Tech Lab Summit made the case that it’s coming to the wires between every console. AAMP isn’t a product, but the connective tissue that determines whether agentic advertising plays out as a handful of closed garden experiences or as one interoperable open ecosystem where the buy and sell sides can still talk to each other at machine speed.

For our clients, execution friction is on its way out. The advantage shifts upstream, to the quality of first-party data, the discipline of structured brand content, the clarity of campaign intent, and the human judgment applied at the portfolio level. Success in an agentic ecosystem calls for machine-readable credibility over traditional channel certifications. To capture demand, brands must ensure autonomous systems can seamlessly understand, trust, and select them on behalf of consumers.

AAMP 3.0 is the next milestone. The summit was a clear signal that the work of building the rails is no longer theoretical. It’s shipping.

Read more about AAMP and the IAB Tech Lab agentic roadmap here.