Skip to main content
Findings Review is where a human reviewer works through the assessment’s findings one at a time, reads the AI’s rationale and its cited source passages, and records a decision. Findings Review, one finding open with its rationale, evidence requirements and a decision panel

​
A finding

Each finding shows:
  • The proposed rating and the run and confidence it came from
  • The decision so far, and how many of the criterion’s evidence requirements have been decided
  • The rationale — why the engine proposed what it did
  • Its evidence requirements — the individual things this criterion checks for, each with its own status (Not found, Partial, or matched) and its own decision
Use Previous / Next to move through the open findings, or Focus on one requirement to work a single evidence requirement at a time. Expand opens the finding full-width.

​
Reading the source

The right-hand Reader panel shows the actual source document, with the passages the engine cited highlighted and numbered against the citations in the rationale. Ask about this document opens a question directly against what’s on screen — see Ask and search. A decided finding, with the source document open on the right and citations highlighted

​
Deciding a finding

At the bottom of the panel:
  • Accept [rating] — agree with the proposed rating.
  • Challenge to… — override the rating and record your own.
  • Query — flag the finding for someone else, with an optional reason.
  • Carry as a condition — the finding travels forward as a condition on the signed record, and onto the next one, rather than blocking sign-off now.
  • Raise action — create a remediation task straight from the finding (see Gaps, Actions and Obligations).
The header always shows the assessment’s live headline: its current position (e.g. Not ready), how many findings are decided out of the total, the submission date, and an Activity log.

​
What’s next