An open protocol for AI-to-authenticated-human handoff

Credentialled professionals — those whose standing can be verified against formal licences, certifications, or regulatory records. Lawyers, property and casualty insurance agents, financial advisors, doctors, engineers, tax planners, accountants, architects.

Established professionals — those whose standing is verifiable through professional body membership, peer recognition, and trusted partner data, even where formal licensing does not apply. PR specialists, management consultants, executive coaches, HR professionals, project managers, and many others.

Every data point in every profile carries its own verification status — registry-verified, partner-verified, VC-verified, or self-declared — returned transparently in every response via a parallel provenance map. AI systems receive granular trust signals and can communicate them clearly to users. The basis of trust is visible, not collapsed into an opaque score.

Trusted partners are organisations that hold reliable, maintained data about professionals and meet ATAH's published partner standards. The types of organisation the protocol works with include regulatory and licensing bodies, professional bodies for credentialled and established fields, review platforms with anti-gaming attestation, and approved independent verifiers.

The model is consistent across partner types. Partners pay ATAH for integration access against a published cost-recovery fee schedule, with waivers and deferred fees for regulators and public-interest bodies. Members gain structured machine-readable presence in AI-mediated environments. The commercial model exists to fund the work of building and maintaining the protocol; it is not the reason the protocol exists.

Three commitments are non-negotiable in the partner model:

• Partner payments do not influence matching ordering or recommendation status.
• Partner data carries equal data-handling and provenance treatment regardless of fee tier.
• No exclusive arrangements within any category — multiple bodies, regulators, verifiers, and review platforms may operate in the same category.

• Not a recommendation engine. ATAH returns provenance-visible candidates based on declared need, verification evidence, category rules, and availability. The user or calling AI platform remains responsible for selection. The non-recommendation status is machine-readable in every response.
• Not a regulator, enforcement body, or complaints adjudicator. Concerns reported through ATAH are routed to the relevant regulatory or professional body.
• Not a payment processor or marketplace. No transaction or commerce capability.
• Not a directory product or AEO replacement. Professionals are not "promoted" or "ranked for visibility" in any commercial sense.
• Not a personal data repository. End-user personal data passes through ATAH only as transient handoff data, held in a transient encrypted vault, and crypto-erased per protocol.
• Not an end-user interface. End-users — whether individuals or businesses — interact with AI systems, which use ATAH on their behalf.

ATAH is the layer above the existing agent-web stack. The protocols below are all machine-to-machine; ATAH is the machine-to-human layer that sits on top of them. It composes with — rather than replaces — these standards. In practice, ATAH is exposed natively as MCP tools (and as REST), so an AI platform already speaking MCP calls ATAH directly with no separate data protocol to wire up.

ATAH does not duplicate the work of identity, credential, or commerce standards. Its specific contribution is the layer above — professional categorisation, the staged handoff lifecycle, the matching engine, the trusted-partner trust model, and the commercial-neutrality and provenance-visibility commitments.