Skip to main content

AI decision auditability

Distinct from the build audit trail: this governs what a shipped product decides when it embeds AI in an insurance decision (e.g. ClaimPilot). Regulators and clients will ask how an automated decision was made — this is the answer.

The controls​

  • EU AI Act risk classification at Definition — classify the feature's risk tier; high-risk features carry extra obligations.
  • Decision logging — inputs, model version, output, confidence, and any human override, logged per decision.
  • Proportionate explainability — a decision that affects a customer can be explained in terms they and a regulator accept.
  • Bias monitoring — for underwriting/claims-adjacent decisions, monitor for disparate outcomes.

Why it's its own standard​

The AI provenance standard records how the code was built. This records how the product's decisions are made and can be justified — a regulatory requirement for AI in financial services, not an engineering nicety. Selected via the NFR catalogue when a feature makes or materially informs a decision.

Standards referenced: EU AI Act, ISO/IEC 42001, FCA Consumer Duty.