With recording enabled, a notetaker bot joins your video interviews, the conversation is transcribed, and the AI produces a structured debrief: an overall score, a hire recommendation, strengths, red flags, and the practical details (salary expectations, notice period) — minutes after the call ends.
Before any bot joins a call, the participants are notified that the interview will be recorded. This isn't a courtesy — it's enforced: the bot will not be scheduled until the recording notice has gone out, and if that check can't be verified, the bot stays out of the call. The bot is also visible in the meeting as a named participant, never a silent listener.
- The bot leaves the call and the transcript is processed
- The AI analyzes the full conversation
- The debrief appears on the interview detail page
- You add your own feedback alongside it — the human read and the AI read live side by side
| Section | What you get |
|---|
| Overall score | 1–10, consistent with the recommendation by rule |
| Recommendation | Strong hire / hire / maybe / no hire |
| Strengths | Specific, evidence-based positives from the transcript |
| Red flags | Typed and severity-rated concerns — e.g. skill exaggeration, salary mismatch, notice period |
| Practical details | Salary expectations, notice period, start date as stated in the call |
Two guardrails keep the debrief honest:
- Critical red flags block hire recommendations. If any red flag is rated critical, the recommendation can't be "hire" or "strong hire" — the inconsistency is structurally impossible.
- Scoring is evidence-only. Communication is scored on clarity of reasoning and relevance — explicitly not on accent, dialect, filler words, or non-native fluency. The analysis is auditable: every claim traces back to the transcript.
When the AI's recommendation is strong hire, the interview surfaces a pre-filled offer prompt so you can move while the momentum is real. It's a prompt, not an action — the offer is only ever created and sent by you.
- Transcribes and analyzes the interview into a scored, evidence-based debrief
- Extracts salary expectations and practical details so nothing said on the call is lost
- Surfaces an offer prompt on a strong-hire read
- Never acts on its own recommendation — hire decisions, offers, and anything involving money are always yours
Failures are visible, never silent:
- Missing transcript — if a transcript never arrives for a recorded interview, detection kicks in and you're alerted, rather than the interview quietly showing no analysis. Stuck bots are recovered automatically where possible.
- Analysis failure — if the AI can't produce a reliable read (an outage, an unusable transcript), you get a clearly flagged "needs human review" state with no score, never a fabricated recommendation. Failed analyses are retried automatically.
- Thin or garbled transcripts — a transcript too short or noisy to trust is flagged with a caution rather than scored as if it were clean.