Speak Your Interview Debrief Before You Forget
Talk through an interview for a few minutes and get scorecard notes that separate what the candidate showed from what you felt.
Interview feedback usually gets written hours later, from memory, as a line like "strong fit, good energy". By the panel meeting the specific answers are gone and candidates are compared on impressions.
That is the gap this drop closes. It is a hiring & hr skill built for ChatGPT, and it takes about ten minutes to set up the first time. After that it runs in under a minute.
Who should use it
Hiring managers and HR officers who run interviews and have to write feedback for the panel.
How it works
The skill file does five things, in this order.
1. Save your scorecard criteria once. Paste the role's criteria and what good looks like into a custom GPT or a saved chat. Every interviewer on the panel then debriefs against the same list.
2. Speak within minutes of finishing. Open ChatGPT voice mode on your phone right after the candidate leaves. Describe the questions you asked and what the candidate actually said for each criterion.
3. Run the prompt on the transcript. Paste the transcript into the debrief tag with the role, stage and rating scale. Use a candidate reference instead of the full name.
4. Read the gaps and flags first. Check which criteria came back "Not assessed" and which comments were flagged as unrelated to the job. These matter more than the suggested ratings.
5. Bring evidence to the panel. Correct anything the tool got wrong, then share the notes before the panel meeting. Use the follow-up questions to plan the next interview stage.
What comes back
A per-criterion table of evidence, suggested ratings marked as suggestions, gaps marked "Not assessed", flagged irrelevant comments and follow-up questions for the next round.
The mistake to avoid
The common mistake is speaking only impressions into the debrief and letting the tool turn them into neat, confident ratings. Describe what the candidate said and did, and treat the output as a note-taking aid, since the hiring decision and its fairness stay with the panel.
Running it
Paste the prompt into the Instructions field of a new custom GPT, or use it directly in a chat. If the task involves files, turn on the code interpreter so calculations are executed rather than estimated.
The prompt
You are helping a hiring manager turn a spoken interview debrief into evidence-based scorecard notes. The hiring decision belongs to the panel, not to you.
Role: {job_title}
Candidate reference: {candidate_id}
Interview stage: {interview_stage}
Rating scale: {rating_scale}
<criteria>
{scorecard_criteria_and_what_good_looks_like}
</criteria>
<debrief_transcript>
{voice_debrief_transcript}
</debrief_transcript>
Do the following:
1. For each criterion, list what the candidate actually said or did, as reported in the debrief. Use the interviewer's own words where possible.
2. Separate evidence from impression. Any opinion with no described example goes under "Impression, no evidence".
3. Suggest a rating only where there is evidence. Label it [Suggested] and justify it in one line.
4. If a criterion was not discussed, write "Not assessed". Never infer a rating from other answers.
5. Flag any comment about age, nationality, religion, marital status, pregnancy, accent or other personal matters unrelated to the job, and keep it out of the notes.
6. List up to three follow-up questions for the next stage that would close the gaps.
7. If the transcript is too short or unclear to complete the notes, say so and stop.
Mark anything you infer rather than read directly with [Inferred]. Do not recommend hire or no hire. Output one table per criterion, then the flags, then the follow-up questions. Reply in the language of the transcript.Skills that pair with this one
Bilingual Interview Guides That Work in Arabic
Build interview scripts that work in both English and Arabic without losing cultural context
Train Your Team to Voice Their Tasks in One Session
Get everyone using the same voice-to-task system without chaos or confusion
Train Your Team to Ask Data Questions
Get consistent analysis across your team without writing a procedures manual
The daily skill, in your inbox
One email each morning with the new drop. Unsubscribe in one click.
