An interview ends and the interviewer has another meeting. The recording exists, but the scorecard is still empty. AI notes can help recover the conversation. They can also turn a transcription mistake into confident feedback if nobody checks the source. A useful Greenhouse Notetaker rollout should make that check easy and give the interviewer clear ownership of the final assessment.
This guide covers the operating workflow around the feature. It is not a promise that every interview should be recorded. Start with the interviews your team can support, establish candidate communication and keep ordinary manual notes available when recording is inappropriate or unavailable.
Separate the conversation from the assessment
Greenhouse Notetaker records and transcribes video interviews, then generates notes and summaries. Greenhouse lists Zoom, Google Meet and Microsoft Teams as supported providers. It does not support in-person or phone interviews. Its outputs give the interviewer material to consult while completing a scorecard.
Keep three records conceptually separate. The transcript represents what the system captured. The AI summary condenses that material. The scorecard contains the interviewer's evaluation against the job criteria. A clear summary does not prove that the candidate demonstrated an attribute. That judgment belongs to the interviewer, with evidence they can explain.
| Material | Use it for | Review before relying on it |
|---|---|---|
| Recording and transcript | Checking the conversation | Speaker attribution, missing context and transcription errors |
| Question notes | Finding evidence for a question | Whether the summary matches the actual response |
| Overall summary | Recalling topics discussed | Whether uncertainty or qualifications were dropped |
| Interviewer scorecard | Recording the human assessment | Reasons tied to the agreed job criteria |
Build the interview kit before adding AI
Greenhouse interview kits support custom questions and links between questions and scorecard attributes. Use that structure to agree what each interview should establish. Write the question, the attribute being assessed and the evidence the interviewer should look for. Avoid asking every interviewer to assess everything.
For example, an implementation interview might ask the candidate to explain how they handled a failed customer data import. The interviewer can look for a clear account of diagnosis, recovery and communication. This is an illustrative question, not a claim about a particular candidate or customer. The final scorecard should distinguish the candidate's description from the interviewer's interpretation.
Configured questions also give the notes a useful organizing structure. Greenhouse documents question notes alongside scorecards and an overall factual summary. Do not use the existence of a generated section as proof that the interviewer actually covered the question. Check the underlying conversation when coverage is uncertain.
Confirm consent and interview eligibility
Before the interview, check that Notetaker is enabled at the relevant levels, that the scheduled interview has supported video conferencing and that the candidate's consent status permits recording. Greenhouse documents consent collection through availability and self-schedule requests, as well as a standalone consent request email. Verify the controls in your own account before training interviewers.
Write a fallback for a candidate who declines or changes their preference. Explain who checks the status, who handles a request during the meeting and how the interviewer continues without recording. Have counsel review the notice and consent approach for your hiring locations. A visible meeting participant is not a substitute for that review.
Give interviewers a short review checklist
After processing, Greenhouse provides recordings, searchable transcripts and generated notes alongside the scorecard. Its documented workflow includes clickable timestamps for checking the source and controls for adding generated material to the scorecard. Use those controls deliberately rather than accepting the whole summary without review.
- Confirm the material belongs to the correct candidate, application and interview.
- Read the question notes and identify claims that will affect the assessment.
- Check those claims in the transcript or recording, including who said them.
- Correct inaccurate details and preserve uncertainty where the evidence is incomplete.
- Add an independent assessment in the scorecard and review it before submitting.
Pay particular attention to technical names, numbers, negation and statements about personal responsibility. A candidate who helped a team deliver a project should not become its sole owner in the summary. If the source is unclear, record that uncertainty instead of asking another model to fill the gap.
Greenhouse says a reminder appears when an interviewer adds generated notes without their own assessment. The reminder does not prevent submission. Your review process therefore needs a real owner. Do not describe that reminder to managers as a hard approval gate.
Keep visibility aligned with independent feedback
Greenhouse offers scorecard visibility options of Never, Always and Only after submitting their own. Choose a policy that fits your interview process. If you want independent assessments before a debrief, review whether interviewers should see other submitted scorecards before writing their own.
Apply the same principle to any added automation. A shared AI digest sent before independent feedback could expose opinions that the scorecard policy was intended to withhold. Review access for recordings, transcripts and notes as well as scorecards. Test with representative user roles rather than checking only from an administrator account.
Plan retention and failure handling separately
Recording retention and retention of transcripts or notes are different concerns. Greenhouse documents automatic recording deletion while transcripts, notes and scorecards are preserved, with separate data retention rule options for Notetaker data. Decide which materials your team needs and test the rules for each type. Avoid exporting copies into shared folders without a deletion owner.
Also test an interview where processing fails or useful notes never appear. The interviewer should still know how to complete a scorecard from their own notes. Keep an issue path for reporting incorrect attribution or summaries. Make the missing-recording state visible to the recruiting coordinator without exposing interview content more widely.
Mistakes to watch for
- Copying the AI summary as the entire assessment.
- Treating a nonblocking reminder as mandatory human approval.
- Using another interviewer's opinion to complete independent feedback.
- Assuming recording deletion removes every transcript and note.
- Leaving candidates or interviewers without a manual fallback.
Pilot the workflow on an approved interview format. Review the accuracy of evidence, interviewer corrections and completion of feedback. Measure review work as well as drafting work. Keep the rollout narrow until the team can explain what happens when consent changes, source material is missing or generated text is wrong.
How Remqa helps with interview feedback workflows
Remqa can help map interview preparation, consent checks, reviewed notes and scorecard handoff into a recruiting workflow. Start with AI agents for recruiting teams, review where manual work sits in recruiting hours, or use the recruiting tool stack guide to plan the surrounding handoffs. Bring an example interview process and the rules your team needs to preserve.