Every discovery call becomes an opportunity map, two minutes after you hang up

Discovery calls hold the best evidence of where automation pays inside a prospect’s business, but the analysis lived in one person’s head. Manual post-call writeups took an hour each, so they got skipped, and every call was diagnosed a little differently. Follow-ups went out thin.

I wrote the methodology down before writing any code: anchor on hours per week, separate actions from decisions, apply the instruction-manual test, four probing questions. Only then did I build a silent bot that joins the meeting through Recall.ai and runs the transcript through that methodology with OpenAI.

A few design choices worth noting: launched from a phone (paste the meeting URL, tap once), no database (per-call JSON on a persistent volume), and the bot never speaks in the meeting. It listens, then reports.

Every call now produces the same structured opportunity map about 1 to 2 minutes after hang-up: a summary, the automation wedge with direct prospect quotes, expansion plays, and questions for the next call. Every report lives at a permanent URL, so deal context never disappears into a notebook.