Interview scorecards
Paused with US hiring until mid-November 2026.
A nightly run that reads each day's interview recordings and writes a ranked scorecard Doc for the next morning's hiring review.
- Why
- My ask on 13 Oct: after every call, check the recording and score everybody on one rubric for tomorrow's review. I could do back-to-back interviews and still walk into the review with evidence, not impressions.
- How long
- Set up 13 Oct 2026; the first Doc was made by hand that day
How I built it
- I record every call in Otter.
- The run matches each calendar event to its recording by date and time, not by title, because titles get garbled.
- It reads the full transcript, not the summary, and works out who is speaking.
- It scores 7 areas from 0 to 5, each with one line of evidence from the call.
- It writes a ranked Google Doc, carrying earlier candidates forward, ready for the morning.
- To copy it you need
- Claude, or the AI of your choice, with Claude Code, an Otter connector and Google Docs. The gotcha: Otter's action items mix up who owes what. Treat them as leads and check the transcript.
Try it yourself
Copy this prompt and paste it into ChatGPT or Claude to try it yourself.
I'm looking at Shweta Kamble's interview scorecards on whatarewebuilding.ai. Every weeknight a scheduled Claude task matches the day's calendar interviews to their Otter recordings by date and time, reads each full transcript, scores 7 areas from 0 to 5 with one line of evidence each, and writes a ranked Google Doc for the morning review. Help me build a starter version for the role I'm hiring. Ask me for the 5 to 7 things that matter most, then write the rubric and the run's instructions. The gotcha Shweta hit: recording titles get garbled, so match by time, never by title. Ask me for anything you need.
Tools I used for this project
