Structured outputs, evidence and freshness

Home · Structured outputs, evidence and freshness

Goal

This page explains how to review an agent structured output before using it in a decision, PM Document, publication or governed action.

Read a structured output

1. Read the Summary.
2. Read Findings and Decisions.
3. Open Evidence and citations.
4. Check Source freshness.
5. Read Confidence and Missing information.
6. Decide whether the output stays exploratory, becomes a PM Document or needs escalation.

Decision table

Situation	Risk	Recommended decision
No evidence	Result cannot be justified	Do not publish; add context or source
stale source	Source is too old	Refresh, reimport or confirm manually
conflicting source	Sources disagree	Human arbitration required
unavailable source	Evidence could not be retrieved	Treat as an alert, not a proof
Low confidence	Result is uncertain	Keep exploratory or request review
External action proposed	Impact outside ProPM Agent	Use Actions and approvals
Shareable deliverable	Versioning and governance needed	Open PM Documents and artifacts

Human review

AI outputs require human review before publication, sponsor decision, external communication, customer notification, ticket creation or governed action. Confidence is not approval.

Downstream flow

The recommended flow is run → structured output → artifact → version → PM Document → Download / Publish / Add to knowledge.

Support and audit IDs

Fields such as Trace ID, Structured output ID and Context snapshot ID are useful for support and audit. They do not replace business evidence review. Use AI Log and Support diagnostics for investigations.