Meeting notes recap. Meeting intelligence explains.

The AI notetaker market has spent years perfecting the recap: transcripts, summaries, action items. Useful — and radically incomplete. Meetings are not information transfers; they are negotiations between people, and the outcome usually turns on things no summary keeps: the tradeoff nobody discussed, the deadline that was 60% shrug, the question that got answered with a different answer. Donna’s two-pass analysis was built for exactly that residue.

Pass one: the forensic extraction

Before a word of the report is written, Donna extracts structured facts from the timestamped transcript: every decision (and whether a tradeoff was acknowledged — its absence is flagged as a red flag), every commitment in exact words with soft deadlines marked, every tension with who stood where, every open question left dangling, every position shift and its pivot point, and the observable power dynamics — who deferred, who interrupted, who had the final word.

Pass two: the report with opinions

Then the report is written the way the sharpest person in your company would write it — first person, blunt, every claim carrying a citation:

Why cloud notetakers don’t do this

Partly product focus — summaries satisfy the demo. But partly custody: analysis this candid about your own people (who dominated, who got talked over, who dodged) is only comfortable when the report lives on your infrastructure, behind your authentication, and nowhere else. Depth and self-hosting aren’t two features of Donna; each is why the other is possible.