AI transcribes your meetings now. Great. But the raw output is a wall of "um"s, tangents, and three people talking over each other — useless the moment you need to act on it. The win isn't the transcript. It's the cleaned version.
The problem with raw transcripts
A verbatim log preserves everything and clarifies nothing. You still spend twenty minutes digging for the one decision and the two action items. That's not saving time — that's moving the work.
What I do instead
- Summarize to decisions. Ask AI: "List only the decisions made, with who owns each." Not the discussion — the outcome.
- Extract action items. "Give me a checklist: task, owner, due date." One line each.
- Humanize the tone. Raw AI notes read like a police report. Run them through a fixed prompt — vary sentence length, cut the filler, make it sound like a person wrote the recap. The prompt pack I use for notes, email, and reports is here: https://felovery.gumroad.com (code FIRST50 = half off).
Verify it reads like you
Before you send the recap to the team, paste it into a free AI-risk scorer: https://humanizeai-lab.pages.dev/tools/ai-risk-scorer.html — aim under 40%. A clean, human-sounding recap gets read. A bot-looking one gets ignored, and the meeting happens again.
What's your current meeting-note workflow — still doing it by hand?
Top comments (0)