Review AI-assisted intake

AI-assisted intake helps structure a large document set, but it never writes directly into the proof graph. Every extracted item enters a review queue first.

The review sequence

1. Add the documents you want to analyze.
2. Wait for extraction to finish and inspect any document-level warnings.
3. Review each proposed fact in context.
4. Edit wording, reject noise, or approve the proposal.
5. Confirm suggested links to claims and elements.

What to check

• Attribution: Does the source actually say this, and who said it?
• Precision: Is the proposal a fact rather than an argument or conclusion?
• Date and sequence: Has the system preserved temporal context?
• Support: Is the cited passage the right anchor for the fact?
• Privilege and sensitivity: Should this material enter the matter workspace at all?

Keep human review meaningful

Batch approval is tempting, but it defeats the reason the queue exists. Review uncertain and high-impact proposals first, then use consistent decisions to move through the remainder efficiently.

After approved extraction has populated the graph, use Case Context to search the assembled records and Case Assistant to ask source-grounded questions or request individually reviewable metadata proposals.