Audit logs
Make every AI request explainable after the fact.
ModelSpend records the evidence developers, finance and governance teams need to understand how AI spend was created and why a route was chosen.
What to log
Request context
Application, workflow, route hint, timestamp and tenant context.
Routing decision
Selected provider, model, routing tier, fallback path and policy influence.
Cost and latency
Estimated or actual cost, token usage, response time and provider status.
Outcome
Successful response, upstream error, validation issue, policy denial or timeout.
Why it matters
Audit logs make AI usage visible. They help teams debug failures, attribute spend, investigate anomalies and prove policy behaviour.
Avoid logging sensitive prompt content unless your tenant configuration, privacy requirements and retention policy explicitly allow it.