AI-vetted hiring: what our screening actually measures

“AI-vetted” has become a sticker. Since we sell it, we owe you the specifics of what actually runs between a candidate applying and a profile landing on your desk.

Stage one: code that resembles the job

Candidates complete a task drawn from real production scenarios — a refactor with failing tests, an API with ambiguous requirements. Models score correctness, code quality, and how the candidate handled the ambiguity. Puzzle-solving trivia is deliberately absent; it predicts interview performance, not job performance.

Stage two: communication under structure

A structured video interview, transcribed and analysed: clarity of explanation, question-asking behaviour, and how the candidate communicates uncertainty. In distributed teams, this predicts success more strongly than any credential. [PLACEHOLDER: cite internal data when available]

Stage three: the human gate

A senior architect reviews every profile that passes stages one and two — with the evidence attached. They reject roughly a third of what the models pass. That ratio is the point: AI creates leverage, humans create accountability.

You see three profiles with assessment evidence, not a stack of keyword-matched CVs. If the shortlist rates below 8/10, we recalibrate and rerun free.