What AI decides

Extracts evidence

The AI identifies role-relevant claims in a candidate's interview and materials, each tied to its source quote. Authority: none on its own — every extraction shows its receipt, so a reviewer can check the reading against the source and discard it.

Scores interview facets

It scores twelve facets across four meta-traits from interview evidence, each with a stated confidence, abstaining where evidence is thin. Authority: advisory. Scores describe the strength of observed evidence, not ability as a proven fact and not future performance.

Orders evidence

For an employer's search, it orders candidates by the presence, specificity, and checkability of role-relevant evidence against the employer's stated criteria. Authority: advisory. Ordering reflects evidence availability and quality — not a prediction of who will perform best — and is fully reviewable.

Establishes a qualified pool

It distinguishes candidates for whom sufficient role-relevant evidence exists to justify serious consideration from those for whom it does not yet. Authority: advisory, and reversible — a human can look at the evidence and disagree.

Flags convergence and divergence

It compares the interview evidence with the candidate's self-report and flags where they corroborate or diverge. Authority: informational. The flag invites a human question; it does not carry a conclusion.

Drafts interpretations

It writes plain-language summaries of what the evidence shows. Authority: none on its own. Interpretations are the index into the evidence, not a substitute for it — each links back to what the candidate actually said.

Abstains

Where the interview did not surface real evidence for a facet, the AI outputs “insufficient signal” rather than inventing a score. This is the one output the system treats as non-negotiable: a gap is reported as a gap.