Agentic advertising is the model where AI agents negotiate, agree, and execute media deals within mandates set by human teams. This page answers the questions buyers, publishers, and marketing teams ask most often.

At a glance

Agentic advertising: a buying model where AI agents act on behalf of humans within pre-defined rules and approval thresholds

Buying agent: software that evaluates inventory, places bids, and commits to deal terms autonomously

DealSheet: the bilateral record of deal terms agreed between a buying agent and a publisher, owned by both parties

AAMP: the IAB Tech Lab Agentic Advertising Management Protocols standard that governs agent communication in advertising

Earned autonomy: a progression model where agents gain greater independence as they demonstrate reliable performance

What is agentic advertising?

Agentic advertising is a form of programmatic buying where AI agents negotiate and execute media deals without requiring human sign-off on every transaction. The human team sets the mandate: budget, audience parameters, viewability floors, brand safety rules. The agent operates within that mandate.

It is not fully autonomous buying. Humans set the boundaries, approve the deals that exceed defined thresholds, and retain oversight at every meaningful decision point. Agentic advertising reduces the time and operational cost of media buying without removing human control.

How does agent-to-agent negotiation work?

A buying agent evaluates available inventory against the mandate it has been given. It contacts a publisher's selling agent, which applies the publisher's own deal rules. The two agents negotiate terms: price, format, placement, floor compliance. If both sides accept, the deal is recorded in a shared deal record.

The negotiation happens at machine speed. What previously required emails, calls, and insertion order sign-offs is reduced to a structured exchange between two systems operating under human-approved rules.

What is the AAMP standard?

AAMP is the Agentic Advertising Management Protocols standard, published by the IAB Tech Lab. It defines how AI agents communicate, structure deal proposals, and exchange governance data across platforms. Find it at the IAB Tech Lab AAMP standards page.

AAMP is the infrastructure layer that makes interoperable agent-to-agent advertising possible. Platforms built on AAMP can communicate with each other regardless of which tools the buyer and publisher use. Alkimi implements AAMP as the communication standard for all agent interactions on its marketplace.

What is a DealSheet?

A DealSheet is the bilateral record of what was agreed between a buying agent and a publisher's agent. Both parties hold the same record. Neither can alter it unilaterally after the deal is committed.

The DealSheet records price, format, placement, brand safety conditions, and any custom deal terms. It serves as the audit trail if there is a dispute, and as the reference point for any downstream activation. Unlike an insertion order, the DealSheet is machine-readable, structurally consistent, and accessible to both parties without intermediary systems.

What is earned autonomy?

Earned autonomy is the governance model that determines how much independence an AI agent is given. It works as a progression through five stages:

Observe: the agent monitors and reports, no action taken
Recommend: the agent suggests actions for human review
Draft: the agent prepares deal terms for human approval before committing
Human-approved action: the agent executes deals that fall within pre-approved parameters
Bounded automatic action: the agent executes within defined limits without per-deal approval

An agent moves up this progression by demonstrating reliable performance at each stage. It does not start at full autonomy. The model is designed to build trust incrementally, with humans retaining the ability to pull back at any stage.

How do human approvals work?

Human approvals are applied at the deal level. Buyers define approval thresholds when they configure their agent's mandate. Deals that fall inside those thresholds can be executed automatically. Deals that exceed them trigger a human review.

This keeps humans in control of material decisions without requiring them to review every individual transaction. The agent handles the high volume, low-variance work. Humans approve the exceptions.

What do publishers need to accept agentic deals?

Publishers need to configure a selling agent with their own deal rules, floor prices, brand category restrictions, and approval workflows. They need to connect to a marketplace that supports AAMP so their agent can communicate with buyer agents on other platforms.

Publishers also benefit from adopting a structured deal record format so that deal terms are machine-readable and auditable. The Alkimi marketplace provides this through its DealSheet infrastructure, which gives publishers a bilateral record of every deal their agent accepts.

Can small advertisers use agentic advertising?

Yes. Agentic advertising does not require a trading desk. A buying agent operating within a well-configured mandate can access premium inventory on behalf of a small advertiser without the operational infrastructure that large buyers typically maintain.

The mandate configuration is the critical step. A well-defined mandate with clear audience parameters, budget controls, and brand safety rules gives a small advertiser's agent the guardrails it needs to operate effectively. Marketplaces like Alkimi are designed to be accessible to advertisers who do not have dedicated programmatic teams.

What are the risks of AI agent media buying?

The main risks are mandate misconfiguration, inventory quality, and auditability. If a buying agent's mandate is poorly defined, it will execute deals that technically comply but do not serve the advertiser's actual objectives. If the marketplace does not provide transparent deal records, the buyer cannot verify what was agreed. If the agent operates without human oversight thresholds, material decisions may go unreviewed.

These risks are manageable. Clear mandate documentation, bilateral deal records, and a tiered approval model address all three. The governance infrastructure matters as much as the agent itself.

How do AI agents handle brand safety?

Brand safety conditions are part of the buying mandate. The agent evaluates every deal opportunity against those conditions before committing. If a placement does not meet the brand safety parameters, the agent declines.

In a structured marketplace, brand safety conditions are also part of the deal record. Both parties have visibility into what was agreed, and any deviation from agreed conditions creates a verifiable record. This is a stronger audit trail than many manual buying workflows provide.

What KPIs should teams track for AI agent media buying?

Track mandate compliance rate (percentage of executed deals that fall within mandate parameters), deal completion rate (percentage of agent negotiations that result in a committed deal), cost-per-completion (how much each successful deal costs in agent API calls and execution time), and audit hit rate (percentage of deals reviewed against deal records after execution).

Standard delivery KPIs (viewability, completion rate, brand safe rate) still apply. The additional KPIs specific to agentic buying measure how well the agent is operating within its mandate, not just whether the media is delivering.

What is the difference between AAMP and A2A?

AAMP is the advertising-specific protocol published by the IAB Tech Lab. It defines the deal structures, governance fields, and communication standards for agent interactions in media buying and selling. A2A (Agent-to-Agent) is a general-purpose protocol for agent communication, developed outside advertising.

In practice, advertising platforms use AAMP for deal-level communication because it includes the fields that advertising transactions require: brand safety conditions, audience parameters, viewability floors, and deal governance data. A2A provides the underlying communication layer. AAMP sits on top with the advertising-specific structure.

How does Alkimi work as an agentic marketplace?

Alkimi is the marketplace where buying agents and selling agents meet to negotiate and commit to deals. It provides the DealSheet infrastructure (the bilateral deal record both parties hold), the governance layer (earned autonomy controls and approval workflows), and the AAMP-compliant communication standard for all agent interactions.

Alkimi does not replace a buyer's planning tools, DSP, or ad server. It handles the part of the workflow those systems were not built for: the negotiation between a buying agent and a publisher's selling agent, and the governance record that makes that negotiation auditable.

FAQs of Agentic Advertising