Interviews
Every tracker schedules the interview. This one reports what happened in it.
The hiring decision is made in a room nobody recorded, and reconstructed afterwards from one person's memory and four sets of notes that disagree.
Of nine recruitment sites we studied, not one has a page about what happens inside the interview.
- Interview · completed
- 32 minutes
- Panel · 2
- Scorecard · submitted
- A claim about owning the migration, at 04:12
- A notice period longer than the brief allows, at 11:48
- An answer that trailed off and was never chased, at 19:03
- A two-part question where half went unanswered
The write-up
Every finding carries the moment it came from. Press it and hear it.
A bot joins the call and records it — with consent captured on the file and a retention date attached to it. What comes back is not a summary. It is a set of findings, each one timestamped: a claim the candidate made, a risk worth knowing about, a question that was answered only halfway.
You do not have to take any of them on trust. Each one seeks the recording to the second it happened, which is the check that matters and the one you should run on the call.
- Interviewer
Tell me about a system you owned end to end.
- Candidate
We moved the reconciliation job off the main database in the second quarter — it was locking the payouts table every night.
- Interviewer
Who decided that, you or your lead?
- Candidate
I proposed it, wrote the migration plan, and ran the cutover on a Saturday with two people from ops.
- Interviewer
And what did it cost you?
- Candidate
Honestly the rollback plan was thin. If it had failed we would have been restoring from a snapshot.
- Interviewer
When could you start?
- Candidate
Realistically I need to serve sixty days, possibly ninety if the handover slips.
A finding whose quote does not appear in the transcript word for word is dropped before you ever see it.
Three things worth knowing
The guarantees, not the machinery.
Nothing is invented. The recording is transcribed, and a finding that cannot be verified against that transcription is dropped before anyone sees it. A fabricated quote inside a hiring decision is worse than no quote at all, so the failure mode is deliberately silence rather than confidence.
The panel is checked too. Each panellist's rating sits beside what the interview itself produced, competency by competency, and the rows where they are furthest apart are marked — before the debrief rather than after the hire.
The interviewer gets read as well as the candidate. Question by question: was it a clear question, did it ask two things at once, and did the answer land, land partially, or go somewhere else entirely. It is labelled in the product as coaching on question craft, because that is what it is — not a signal about the person being interviewed.
- Consent and retention on the file —
Not a policy someone remembers. A date attached to the recording.
- Recording is optional per interview —
Switchable at scheduling, and again at reschedule.
- Upload works too —
A recording made elsewhere can be attached and read the same way. Worth asking us how retention is handled on an uploaded file specifically — the answer is not identical to the bot's.
- It shares its foundation with call intelligence —
The same recording and transcription layer runs your sales calls. What sits on top is deliberately different: an interview is read for claims, risks and question craft, not for objections and buying signals.
Before you buy it for this
What it does, and one thing it does not.
›Can we read the full transcript on the interview?
Not today. The recording is transcribed and every finding is verified against that transcription, and each finding plays the moment it came from — but the transcript is not a screen you can open and scroll on the interview. If reading it end to end matters to you, say so on the call rather than assuming it is there.
›Does the candidate have to consent?
Consent is captured on the recording, and recording can be switched off for any interview at the point of scheduling. The compliance screens tell you what is held, for whom, and until when.
›Which platforms does the bot join?
Google Meet and Microsoft Teams. A recording made elsewhere can be uploaded and read the same way.
›Does it score the candidate for us?
It suggests competency ratings with the evidence attached, and it flags risks. It does not advance, reject or decide. Where it cannot support a finding, it drops the finding.
›What about interviewer bias?
Two mechanisms, and neither is a judgement. A rating far from what the interview produced gets surfaced for a look. And the question-craft read is about the interviewer's half of the hour, which is the half nobody else reports.
›How long are recordings kept?
For the retention window you set on the recording. Ask us to show you the date on a real file rather than the setting on a policy page.
Book a demo
Make us play the moment.
This is the section of the demo worth booking for. Bring a recording if you have one — and pick the finding you least believe.
- Play a recorded interview and open the write-up beside it
- Press a finding and check it against what you hear
- Open a panel divergence and read both readings
- Open the question-craft read and find a two-part question
- Switch recording off for an interview and see what changes
- Open the retention date on a recording