What we mean by agent observability

Agent observability is the record of what an AI agent actually did on the job: what it was asked, what context it retrieved, which tools it called, and what happened to its output afterwards.

The detail

It is not model evaluation. Evals tell you how a model performs on a benchmark. Observability tells you how a specific agent performed on Tuesday, in production, for a named team.

The distinction matters because almost every organisation we speak to has the first and none of the second.