Purpose
What task is the system expected to support, and for whom?
PUBLIC RESOURCE · EARLY STAGE
AI evaluation guide
A short introduction to making review conversations more structured. This guide does not publish internal scoring methods or imply that an AI system has been independently evaluated.
Questions that help a review
Evidence
What observable result would help a reviewer understand what happened?
Limits
What should the system decline, escalate or leave to a person?
Recovery
How can a team notice a failure and correct the next step?
These prompts are discussion aids. They are not a score, certification, human-expert review or readiness decision.