Audit Trail
An audit trail is a durable, reconstructable record of what an AI system did — the inputs, decisions, tool calls, and outputs — so you can later explain or investigate any action it took. It's what turns 'the agent did something' into 'here's exactly what it did and why.'
When an AI agent takes real actions, you need to be able to answer, after the fact, what it did and on what basis. An audit trail is that record: each decision, the context it had, the tools it called and with what arguments, and the output it produced — logged durably enough to reconstruct an incident weeks later or show a regulator.
It’s the backbone of AI governance, and it does double duty. Operationally, it’s how you debug a failure and trace where things went wrong. For compliance, it’s the evidence that the system stayed within policy. And strategically, it’s what lets an organization trust an agent enough to widen its scope over time — you can grant more autonomy when you can always see what was done with it. The overlap with observability is real: observability gives you the live traces, and the audit trail is the retained, governance-grade record built from them.