How Hireflo scores an interview

People ask “how does the AI score?” as if there is a secret number at the end.

There is not. Hireflo scoring is a short loop you can draw on a whiteboard. If a step cannot be shown, we do not ship it.

1. The job becomes checks

You pick a role or paste a job description. Hireflo writes binary criteria — not “strong communicator,” but “explained a tradeoff in their own words,” not “senior,” but “named who owned the incident and how they would page.”

The point of a check is that two interviewers can disagree about whether it matters and still agree about whether it happened. That is the debrief you want.

The question library is a starting shelf. A pasted JD still gets a fresh rubric. An AI Engineer does not inherit a Forward Deployed Engineer’s room. Hiring flow is per role on purpose.

2. The room runs the list

A speech model converses from the rubric and the question plan you configured. It asks follow-ups. It probes a slogan. It does not fill silence for the candidate — that is the interview.

Question tracking checks topics off as they get covered. The candidate is in a live session (WebRTC) when the processor is enabled; otherwise the product fails closed into a question-card mode. Consent happens before a camera. That order is not a UX flourish. It is a rule.

3. Credit is gated on the tape

After the session, an assessment model scores each criterion as pass, miss, or insufficient evidence.

A pass is refused unless:

• the supporting quote appears in a candidate turn, and
• the report can show the timestamp.

If the model cannot cite the span, the criterion is not credited. Insufficient evidence is not a 6. It is a refusal to invent a score.

This is the opposite of a 1–10 interviewer rating. The long version is A score without a timestamp is a guess.

4. A person still decides

The scorecard is a recommendation. Your team hires.

You can override a criterion. You can throw out a bad question. You can decide a miss on a nice-to-have is not a miss on the role. You cannot hide behind “the AI said 7.4.” There is no 7.4.

Hireflo also will not auto-reject. If you need that sentence in isolation: Hireflo is not an auto-reject engine.

What never enters the score

• Faces, clothing, the room, the background
• Accent, pitch, tone, or “confidence”
• A single gut number presented as the decision

Those are easy for software to fake-measure and impossible to defend. We do not collect them as evidence because they are not evidence of the work. The public version of this list is the AI Explainability Statement.

Try it

Run a session against a real job. Open the report. Click a score. If the clip is not there, the score should not be there either.

See pricing · Visit the Trust Center · Read the principles