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.

Runtime decisions, not policy documents. Each proposed agent step receives a decision before it runs, from the rules your team set, and the SDK in your application enforces it. The record shows the step, the rule that matched, and any approval, so a reviewer reads what happened rather than what a policy said should happen.

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

Start with one task and follow review to release. Orlo provides operational evidence; your organization retains responsibility for its acceptance criteria, policies, and legal decisions.