# Run final round debriefs from one Lever scorecard matrix

> Open one room before every hiring decision meeting: every interviewer's scores side by side, disagreement highlighted, and a draft debrief ready to file.

- Workflow type: app
- Services: Lever, Slack Bot
- Categories: HR & People
- Published: 2026-08-31

## What it does

- Pick any candidate sitting at onsite or final stage and get one room with everything the panel needs: the role, the resume, prior notes, and who still owes feedback.
- Every interviewer gets a column and every competency gets a row, so the places your panel disagrees show up on screen instead of staying buried in separate forms.
- Nobody sees the matrix until their own feedback is in, so no one anchors on a colleague's score before locking their own.
- One button drafts an evidence based debrief you can edit in the meeting, then files it onto the candidate profile, tags the outcome, and posts it to your panel channel.

## What you'll need

- A Lever account where your panels submit interview feedback forms
- A Slack workspace with a channel for your hiring panels
- Candidates already at an onsite or final stage in Lever, with their interviews scheduled
- Your interviewers' Lever accounts using the same email addresses they sign in to the app with, so it knows whose feedback is whose

## Prompt

Build me an interview debrief room that my hiring panels open before every final round decision meeting. It runs on Lever and messages the panel through Slack Bot. The whole point of the app is that all of the panel's scores sit side by side on one screen, so disagreement is visible in the room instead of buried in separate feedback forms.

The first screen is a candidate picker. Use List Opportunities filtered by stage to show only candidates currently at an onsite or final stage, with the candidate name, the role they are interviewing for, their current stage, and how long they have been sitting in it. Use List Postings to resolve each opportunity's posting into a readable role title and team. Let me search the list by name and filter by posting. Picking a candidate opens their debrief room.

At the top of the debrief room, show the context the panel needs in the meeting. The candidate and the role come from Retrieve Opportunity and List Postings, including the posting's requirements so the panel can score against what the role actually asks for. The resume comes from List Resumes on an Opportunity, rendered inline or as a link, using the parsed resume data where it is available. Prior notes come from List Notes on an Opportunity in reverse chronological order, collapsed by default so they do not push the matrix off screen.

Next to that, show a feedback coverage strip. Use List Interviews on an Opportunity to get the scheduled interview slots and their interviewers, and List Feedback on an Opportunity to get the forms actually submitted. Anyone who sat an interview but has not submitted a form is listed as still owing feedback, with the date of their interview, so the panel sees the gaps before the conversation starts.

The center of the room is the scorecard matrix: one column per interviewer, one row per competency, each interviewer's score in the cell. Lever feedback forms come back as ordered field sets, so build the rows by aligning fields across forms that came from the same feedback template, where fields in the same position with the same title are the same competency. If a candidate's panel used more than one template, group the matrix by template rather than forcing unrelated fields into a single grid. Under each score, show that interviewer's written comment for the field, expandable, because the comment is the evidence the panel will actually argue over.

Compute the score spread on every competency row and highlight the rows where the panel diverges, so the meeting starts on the rows that need discussion rather than walking the list in order. Show the range and the average on each row. Make the divergence threshold a setting rather than hard coding it.

Bake in the anti anchoring rule: hide the matrix from an interviewer until their own feedback is submitted. Match the person viewing the app to their Lever user with List Users on their email address. If that person has an interview slot on this opportunity and no submitted feedback form, show them a locked state explaining that the matrix unlocks once their own feedback is in, with a link to submit it in Lever, and show them nothing of their colleagues' scores or comments. Everyone else, including recruiters and hiring managers who do not owe feedback on this candidate, sees the full matrix immediately.

Never display diversity survey responses or EEO data anywhere in this app, and never pull them into the debrief draft. This is a decision meeting surface and that data has no place in it.

Put a "Draft the debrief" button in the room that starts a background agent. The agent reads every submitted feedback form from List Feedback on an Opportunity, the interview slots from List Interviews on an Opportunity, the notes from List Notes on an Opportunity, and the job posting requirements from List Postings, then writes an evidence based summary in three parts: what the panel agrees on, exactly where they disagree and what each side cited as evidence, and the open questions the panel never got answered. Every claim should point back to the interviewer who made it. The agent summarizes evidence and never makes the hire call: no hire or no hire recommendation, no manufactured consensus that is not in the forms, and missing panelists flagged rather than silently skipped. Its draft lands back in the app in an editable panel so the panel can rewrite it live during the meeting.

Once the panel is happy with the draft, one click files it. Post the final text onto the candidate profile with Create Note on an Opportunity, passing perform_as so the note is attributed to the real person filing it rather than to the integration. Tag the outcome with Update Opportunity Tags using the outcome the panel picked. Then send the summary and the outcome to the hiring panel's Slack channel with Slack Bot Send a Message, including a link back to the candidate in Lever. Let me choose the channel per posting and have the app remember it.

Remember that Lever's model is one Opportunity per candidate per posting, so everything in the room is scoped to a single opportunity. Keep the room readable on a screen share, since this gets opened live in a meeting: the matrix should stay legible at a glance, with the candidate context available but never dominating the screen.

## How to customize

- Choose which stages count as a final round, so the candidate list matches how your own pipeline is named.
- Set how big a score gap counts as disagreement, so only the real splits get highlighted.
- Pick the outcome tags your panel can apply and which Slack channel each role reports into.

## FAQ

### Does the app decide whether to hire the candidate?

No. It lays out what your interviewers actually said and scored, and the written draft sticks to summarizing that evidence: where the panel agrees, where it splits, and what is still unanswered. It never recommends hire or no hire. The decision stays with the people in the room.

### Can an interviewer see everyone else's scores before submitting their own?

No. If someone was on the interview panel and has not submitted their feedback yet, the matrix stays locked for them and they see a prompt to go submit it first. Recruiters and hiring managers who do not owe feedback on that candidate see the full matrix straight away.

### Does it show diversity survey or EEO answers?

Never. That data is not displayed anywhere in the app and is not fed into the drafted debrief. This is a decision meeting screen and that information has no place in it.

### What if some panelists still have not submitted feedback when we meet?

The top of the room lists exactly who is missing and which interview they sat on, so you can chase them or decide to proceed. You can still draft the debrief from the feedback that is in, and the missing voices are flagged rather than quietly ignored.

### Who does the filed debrief show up as in Lever?

The note is attributed to the real person who filed it, not to a generic integration account, so your candidate profile and audit trail read the way they normally would.

### Will this work if different interviewers used different feedback forms?

Yes. Competencies are lined up across forms built from the same template, and when a panel used more than one template the room groups the matrix by template instead of forcing unrelated questions into one grid.

Use this prompt in General Input: https://www.generalinput.com/prompts/run-final-round-debriefs-from-one-lever-scorecard-matrix