---
title: "Evidence and governance"
description: "Connect your team’s AI decisions to the tests, releases, permissions, and reviews behind them."
order: 0
---

# Evidence and governance

Orlo gives teams a shared record of AI work: what was tested, which version answered, which agent actions were permitted, and what people reviewed. Use that record to decide whether a task is ready, investigate a result, and improve the next release.

| Question | Where to look |
| --- | --- |
| Why was this model selected? | The comparison’s examples, returned answers, scores, and release link |
| What produced this answer? | Its task version, release, model, validation result, and retrieved passages |
| Why did this agent action run or stop? | The submitted step, permission decision, matching rules, and approval |
| Which shared instructions were available? | The skill version, publication history, task pins, and app retrievals |
| What changed after review? | Corrections, saved steps, dataset lineage, and later comparisons |

## Guides

:::cards
- [Read the evidence](/governance/operational-evidence) Understand what each record establishes and who can access it.
- [Checks and permissions](/governance/production-checks) Apply answer rules, agent permissions, and human approval to a workflow.
- [Organize your team’s review](/governance/internal-ai-operating-model) Assign responsibilities and connect review to releases.
- [EU AI Act evidence](/governance/eu-ai-act-operational-readiness) Use operational records as inputs to your organization’s readiness work.
:::

Start with one task and follow [review to release](/console/tasks/improve-from-production). Orlo provides operational evidence; your organization retains responsibility for its acceptance criteria, policies, and legal decisions.