Audit Trail

An audit trail is the durable record of what an AI system did: which action, in which system, on whose authority, and whether a person approved it first.

Once an AI system is changing records and sending messages rather than answering questions, the question stops being whether the output was good and becomes whether the action was authorised. That is not answerable from a chat log. It needs a per-action record that survives independently of the conversation that produced it.

What makes a trail useful is that it captures authority as well as activity. Knowing an agent updated an opportunity is half the answer; knowing it acted as a named person who held that permission at that moment is the half that satisfies a security review. Trails that record only the action tend to fail the first time someone asks a pointed question about one.

Frequently Asked Questions

The action taken, the system it touched, the identity it acted for, the permissions that allowed it, the time, and any approval decision that released it.