An AI inventory tells you what exists. A use-case record explains why a system exists, what it is allowed to do, and the evidence behind the decision to use it.
Record the purpose and owner
Document the business problem, intended users, accountable owner, affected process, and the decisions or actions the system may influence.
Describe data and dependencies
Record important data categories, connected systems, external providers, permissions, retention assumptions, and restrictions on sensitive information.
Capture testing and limitations
Keep the acceptance criteria, representative tests, known failure modes, unresolved limitations, and the conditions under which the system should not be used.
Define human oversight
State which actions require approval, who reviews exceptions, what authority reviewers have, and how users can report unexpected behavior.
Set review triggers
Revisit the record when the use case, model, data, permissions, vendor, affected population, or legal and operating context changes.
Related knowledge
Use this record alongside AI risk classification, human-review models, internal AI policy, and vendor due diligence.
References
This is operational governance guidance, not a determination of legal compliance or suitability for regulated professional decisions.