EVIDENCE OS / AUDIT TRAIL
See what each agent did and
where every number came from.
Evidence OS keeps a record of every workflow run: what was asked, which data was used, what the agents did and what they produced. A reviewer can check a result without asking the person who ran it.
What is an AI agent audit trail?
An AI agent audit trail is the record of what an agent did: the request it received, the data it used, the tools and calculations it ran and the result it produced. It lets someone who was not involved check, repeat or challenge the result.
THE RUN RECORD
What Evidence OS records for each run.
| Record | What it contains |
|---|---|
| Run inputs | The workflow, the request and the settings it ran with |
| Data sources | The sources used, with the references needed to find the same data again |
| Steps | The progress of each workflow step and the agent assignments within it |
| Calculations | The calculations behind each finding, so a reviewer can check them |
| Results | The structured results and the deliverables produced |
| Limits | The tools and budget each agent assignment was allowed to use |
Who is accountable for what.
Evidence OS records what the platform did. Your team decides who owns each workflow and who approves its results. Each project sets out these responsibilities before release, along with how long records are kept.
What the audit trail does not do.
A record shows what happened. It does not show that a conclusion is right. A reviewer still has to check that the data supports the finding.
How to check AI research