Read the AI evaluation

Where: open a candidate in a job → Evaluation.

This is where you spend your reading time. The point isn't the headline rating — it's the evidence underneath it, so you can sanity-check the Assistant against what you see and walk into an interview already knowing what to probe.

The evaluation section shows:

Top-level recommendation — one of:

• Schedule interview
• Schedule exploratory call
• Request clarifications
• Review for rejection

…with a confidence level: High, Medium, or Low (shown as Recommended action (with … confidence)). Hover the recommendation to read the full AI summary.

Scorecard — one row per qualifier. Each row shows:

• The qualifier title (hover the ⓘ for the full criterion text).
• An EvaluationIndicator bar across five levels:
  • Poor match — red
  • Partial match — yellow
  • Clear match — green
  • Exceptional — bright green
  • Insufficient data — grey
• The AI's rating is shown as an AI icon at the appropriate level.
• Recruiter ratings (from comments with evaluations) show as team member avatars at the level they chose. Hover an avatar to read that person's note.
• If Show AI summary is checked, the AI's reasoning appears as a card under the qualifier.

Strengths — bullet points of what the AI highlights as the candidate's strong points.

Clarifications needed — bullet points where the AI flagged gaps or questions.

Data from toggle — switch between Current stage (feedback from the current stage only) and All stages (all recruiter feedback across the whole process).

How to read it well

• Start with confidence, then the evidence. A High-confidence Schedule interview on a full profile is a strong signal; the same recommendation at Low confidence usually means thin data, not a weak candidate — read the summary to see what's missing.
• **Turn on Show AI summary when a rating surprises you.** The reasoning card under a qualifier is the receipt — it tells you why the Assistant landed where it did, so you can agree, or override with a note.
• **Treat Insufficient data as a question, not a knock.** Grey means the Assistant couldn't find evidence either way. That's exactly what the interview, or a quick clarifying message, is for — not a reason to pass.
• **Mine Clarifications needed for your interview.** Those bullets are the open questions about fit — bringing them into the conversation is the fastest way to resolve a borderline call.
• **Use the Data from toggle to control noise. Current stage keeps the picture focused on where the candidate is now; flip to All stages** when you want the full history of what every reviewer has said.

Good to know. Where the AI icon and a teammate's avatar sit on different levels for the same qualifier, that's a healthy disagreement worth a look — hover both to compare the evidence. The team's rating is what counts; the AI is a second opinion, not the referee.