Overview
Answer QA checks AI answers, public data, published outputs, protocols, and customer action paths against approved Commerce Graph data.
- Wrong facts become issues with severity, confidence, proof, and recommended fixes.
- Publishing and protocol checks confirm that public outputs still match approved graph data.
- Retesting closes the loop after fixes.
What Answer QA means in Akii
Answer QA is documented as part of Akii v2 only: Commerce Graph, Agent Gateway, Action Registry, Protocol Compatibility, Answer QA, monitoring, sources, fixes, and Partner Console operations. The page explains the operational role, not a generic marketing claim.
How it works
Akii checks answer accuracy, public data consistency, output health, MCP-compatible output, protocol status, and customer action paths. Issues feed Fix Queue so operators or agencies can repair and retest.
- Start from approved client data instead of scraped fragments.
- Keep public and private data boundaries explicit.
- Attach evidence, owner, status, and version context where the product supports it.
- Retest after changes so issues and readiness gaps do not become stale.
Boundaries
Akii documents what is supported, partial, readiness-only, export-only, planned, or watchlist. Do not treat readiness as certification or exports as full partner integration.
- AI answers are not deterministic; Answer QA records evidence and confidence.
- A passing check does not guarantee future platform behavior.
- A failed action path should be fixed before routing customers there.