An RFP for an agentic advertising platform should test accountability before capability. A demo shows what an agent can do; the RFP has to establish what the agent is allowed to do, who approves what it cannot, which standards it implements at which version, and what record of each deal both parties can retrieve afterwards.

TL;DR. The questions that separate agentic platforms are about governance, not features. Ask how the mandate is set and enforced, which AAMP components the platform uses and at which versions, whether its agent is listed in the IAB Tech Lab Agent Registry, and which deal types it can negotiate rather than only execute. Ask who holds the record of agreed terms and whether the counterparty holds the same one. Finish with data handling, audit access and exit, because a platform that cannot hand back its records cleanly creates a dependency the buyer did not choose.

Why does an agentic platform need a different RFP?

A standard programmatic RFP assumes people make the commercial decisions and software executes them. An agentic platform reverses part of that: software negotiates and commits within limits people set. The questions therefore shift from reach, formats and reporting towards authority, escalation and evidence. Pricing and commercial terms belong in a separate section; the questions here decide whether the platform can be governed at all.

What should an RFP ask about mandates and approvals?

The mandate is the written scope of an agent's authority: what it may buy or sell, at what price, under which terms, and what must go to a person.

How is a mandate defined, and who in the client organisation approves it before the agent acts?
Can the mandate set minimum acceptable terms, such as viewability floors or make-good clauses, as well as maximum prices?
What happens when a proposed action exceeds the mandate: does the agent stop, escalate or proceed?
Can approval thresholds be set by deal value, deal type, seller and data use, and changed without vendor involvement?
How are approvals recorded, and can they be shown per deal?
How does the client pause or revoke an agent's authority, and how quickly does it take effect?

What should an RFP ask about standards support and versions?

AAMP, IAB Tech Lab's Agentic Advertising Management Protocols, is a framework of separately versioned components, not a single specification, and IAB Tech Lab's published material describes no "AAMP compliance" certification. A claim of support therefore needs detail.

Which AAMP components does the platform use (for example agentic-direct, the reference buyer or seller agent, OpenProposal, the Agentic RTB Framework or agentic-audiences), and at which version of each?
How are component version updates tested and communicated to clients?
Can the platform read and produce OpenProposal responses?
Does the platform support AdCP, the separate Ad Context Protocol from AgenticAdvertising.org, and if so at which version and over which transport, MCP or A2A?
Which existing standards does it still rely on, such as OpenRTB, ads.txt (2017) and sellers.json (2019)?

What should an RFP ask about Agent Registry listing and identity?

IAB Tech Lab runs an Agent Registry where companies list the MCP and A2A agents they bring to market, each tied to the registering company's GPP ID and reviewed for validation information.

Is the platform's agent listed in the IAB Tech Lab Agent Registry, and under which company?
How does the platform verify that a counterparty agent is authorised by the advertiser or publisher it claims to represent?
What happens when verification fails: is the agent blocked, flagged or allowed to proceed?

Which deal types should an RFP ask about?

According to IAB Tech Lab, AAMP's support for the Deals API covers programmatic guaranteed, preferred deals, private marketplaces, the open marketplace and linear GRPs.

Which of these deal types can the agent negotiate, and which can it only execute once a person has agreed terms?
Which terms can the agent negotiate within each deal type: price, volume, flight, viewability, frequency, make-goods?
How are mid-flight amendments handled, and does each one need fresh approval?

What should an RFP ask about the deal record?

The deal record is the statement of agreed terms that delivery, invoices, make-goods and disputes are judged against. When each side's agent keeps its own log, two versions of the agreement exist from the first minute.

Where are agreed terms stored, and is the record written at the moment of agreement or reconstructed later?
Does the counterparty hold the same record, and can either side retrieve it without asking the other?
Is each amendment a new, timestamped version, or does it overwrite the original?
Does every invoice line reference a deal identifier that matches the record?

What should an RFP ask about data handling?

Agents exchange briefs, audiences and performance data with counterparties automatically.

What client data does the agent send to counterparty agents, and can the client restrict it?
Under UK GDPR, who is controller and who is processor for each data flow, and where is that set out?
Is client data used to train models that serve other clients?
Where is data stored, for how long, and which sub-processors handle it?

What should an RFP ask about audit?

The case for audit access is long established. The ISBA/PwC study of May 2020 found that only 12% of impressions could be matched end to end across the supply chain. Agentic buying adds decisions to that chain, and each one needs evidence.

Can the client export a full decision log for each deal: offers made, offers received, escalations and approvals?
Can a third-party auditor access logs and deal records directly?
How long are logs and records retained, and in what format?
How are mandate breaches detected and reported to the client?

What should an RFP ask about exit?

Exit questions are easiest to answer before signature and hardest after.

On termination, what happens to deals in flight?
Can the client export every deal record, decision log and approval in a documented, machine-readable format?
Do deal records stay retrievable by the counterparty after the client leaves?
How is the client's agent removed from the platform, and how is its authority revoked with counterparties?

How do vendors answer these questions?

Strong answers name components, versions and records rather than intentions. Alkimi, a neutral agentic advertising marketplace, answers the standards and identity questions within IAB Tech Lab's AAMP framework: its agent is listed in the Agent Registry, it verifies that connecting agents are authorised by the buyer or publisher they represent, and each DealSheet is written at agreement, timestamped and retrievable by either party, with humans keeping approval rights at the deal level.

Frequently asked questions

What is the most important question in an agentic platform RFP?

Who holds the record of agreed terms, and whether the counterparty holds the same one. Delivery, invoices and disputes are all judged against it.

Is "AAMP compliant" a meaningful answer?

Not on its own. IAB Tech Lab's published material describes no AAMP compliance certification, so ask which components a platform uses, at which version, and whether its agent is listed in the Agent Registry.

Should an RFP ask about AdCP as well as AAMP?

Yes, if the buyer works with counterparties that use it. AdCP is a separate protocol from AgenticAdvertising.org, so ask which version and transport a platform supports.

Why include exit questions?

Because deal records and decision logs are evidence the client may need after the contract ends, and a platform that cannot export them leaves disputes unresolvable.

What Should an RFP for an Agentic Advertising Platform Include?