Founder note

Meeting brain: a queryable decision log

A privacy-aware meeting SOP: capture → structure into Decisions, Actions, Open questions → query read-only. Stop re-explaining decisions every week.

Meetings keep producing the same explanations because knowledge dies in the call. Transcripts sit unread, or get pasted into chats with no shape. “AI meeting notes” without a retrieval shape become another unread doc — and privacy risk rises when full transcripts land in shared threads.

What this is

A privacy-aware meeting-brain SOP with three buckets only:

1. Decisions — what we locked, with owner, dated
2. Actions — who / when, blockers flagged, no fluff
3. Open questions — unresolved, dated, with a revisit trigger

Flow: Capture → Structure → Query

• Capture in one place (e.g. a notetaker)
• Structure into the three buckets
• Connect read-only into Claude (connectors / MCP)
• Ask: “What did we decide on X?” without re-explaining

Why it matters

Claude can answer from structured notes. It cannot invent decisions you never wrote down. Structure beats dumping transcripts.

Official Claude connectors make the read path real. Keep the privacy rule: read-only first, write later. Do not dump full transcripts into shared chats. This SOP does not replace legal hold / retention policy.

Humoud’s take

Don’t build a second wiki. Build a queryable decision log. If a meeting doesn’t produce at least one Decision or Action line, it was a status chat — label it that way and move on.

Do this week

1. Template your last 3 meetings into the three buckets.
2. Turn on one read-only connector.
3. Ask Claude two real questions from last week’s decisions.
4. Grant write access only after a clean week of reads.

Source

Adapted from LinkedIn original 2026-09-19-LINKEDIN-HOMOUD-001 (meeting-brain Decisions/Actions/Open-questions SOP). Expanded for durable reference — not a caption paste.

Also see: Automate the routine. Preserve the practice. · Public work