How AI screening works

When candidate assistance is on for a candidate, the Assistant:

1. Reads everything it can find about the candidate for this job.
2. Scores them against each qualifier in the job's scorecard.
3. Produces a top-level Recommendation (with a confidence level) and a narrative summary.
4. Surfaces Strengths and Clarifications needed for your team.

All of this lives in the candidate's Evaluation tab (inside a job's pipeline) or in the Evaluation section when you open a saved candidate against a specific job.

The AI takes a few minutes to run; it can take up to 24 hours in rare cases. You can track its progress — the profile shows "AI evaluation is running…" or "AI evaluation scheduled" until results are ready.

What makes a good evaluation

The screen is only as good as two things: the scorecard and the data.

• A vague scorecard produces vague scores. Qualifiers written as observable criteria ("5+ years managing a backend team") give the Assistant something concrete to match against; "good communicator" gives it almost nothing. Set the scorecard up in the job before you screen — see Build a job.
• A thin profile produces Insufficient data, not a low score. The more the Assistant can see — CV, application answers, recommendations, your team's comments — the more confident the result. That's also why the recommendation can sharpen after you add notes or a candidate replies.

AI screening is powered by candidate assistance (part of The Assistant). Each evaluation consumes 1 token from your plan's AI credit balance.

Good to know. The evaluation is a starting point, not a verdict. Treat Review for rejection as "look here first," not an auto-reject — the Assistant flags fit against your criteria, but it doesn't see everything you know about the role. You stay in control.