What this tool does
The Action Items Extractor reads meeting notes, email threads, or transcripts and returns a clean bulleted list of action items. Each bullet has three parts: who owns it, what they're doing, and when it's due. The structure exists because the gap between "we had a meeting" and "we know what's next" is where most projects quietly stall.
The output format
Each action item follows the same shape:
- WHO — WHAT — WHEN
For example:
- Daniel — finalize the Q3 deck — by Tuesday - Maria — confirm vendor pricing with finance — (no date) - Owner: TBD — push announcement on Twitter — Aug 15
Missing parts are flagged clearly. Owner: TBD and (no date) tell you exactly which decisions need a follow-up question before this can ship.
What makes a good input
- Clear language in the source. "Daniel will send the slides by Friday" produces a clean bullet. "Someone should probably look at this" produces a vague bullet — and surfacing that is doing you a favor.
- Time stamps and names. Transcripts with speaker attribution and timestamps produce much better assignment-to-person accuracy.
- Focused scope. A 30-minute meeting transcript works well. A six-hour off-site transcript needs chunking (paste 30-minute sections separately).
What doesn't work as well
- Brainstorming sessions — the tool will correctly return "no action items found" rather than invent tasks. That's the right answer.
- Transcripts with bad speaker attribution — if the transcript misattributes who said what, action items get assigned to the wrong person.
- Notes with implicit commitments — "we agreed it'd be good if pricing got reviewed" might or might not get picked up depending on phrasing.
Privacy note
Pasted text passes through the AI worker. We don't store the plaintext — there's a 7-day content-hash cache (hash only, not text) so identical pastes return identical extracts free. We don't sell, share, or train on your meeting notes.