Skip to main content

COVP Examples

Fraud Detection AI Product​

A Fraud Detection AI Product is used for analyst triage.

Observed during a 30-day window:

  • 2.4 million scored transactions;
  • 430 active analyst consumers;
  • triage time reduced by 18%;
  • false-positive review burden reduced by 9%;
  • fitness assessment: FIT for analyst triage;
  • value attribution confidence: MEDIUM because process changes occurred simultaneously.

None of these observations changes:

APT = Primary
IC = Detection + Classification
RT = existing evidence-backed tier

They provide contextual evidence about consumption and realized value.

New autonomous use request​

A downstream team proposes using the same Product to block transactions automatically.

Existing COVP evidence says:

FIT for analyst triage
INSUFFICIENT_EVIDENCE for autonomous blocking

The new use therefore requires a new fitness assessment and may change Agency, Intelligence Effect, risk and governance obligations for the new Product/deployment context. It is not legitimate to infer fitness from the previous use case.

High adoption but weak value​

A conversational AI Product is widely used but measured task-completion improvement is negligible.

COVP may trigger a Stage 1 review of Product intent/value proposition. High adoption does not protect the Product from value scrutiny, and weak value does not retroactively make it non-AI.