After a meeting, the notes may contain “we could try this,” “check next week,” and a colleague's name. A polished summary can quietly turn those fragments into commitments that nobody actually approved. A useful decision log keeps the uncertainty visible.
1. Preserve the source
Keep the original notes, meeting date, participants and file version. Before using AI, remove material that your organization does not permit in the selected account or workspace. If processing is not allowed, prepare a neutral summary manually. Do not ask a model to reconstruct missing discussion.
2. Assign a status to every statement
Use four statuses: decided, proposed, open question, and action. A decision needs explicit support in the notes. If an action has no confirmed owner, record “owner not stated” instead of guessing from a name nearby.
You can use Yasnora AI chat to structure permitted text with this prompt:
Using only the notes below, create a table with topic, status (decided / proposed / open question / action), supporting phrase, owner, due date, and what needs confirmation. Do not turn a proposal into a decision. Write “not stated” when an owner or due date is missing. Add no outside facts.
This produces a draft, not an approved record. File and Knowledge features depend on account access, limits and workspace rights; a permitted text excerpt can also be entered directly into chat.
3. Check ambiguous rows
Suppose the note says: “Let's try sending the report on Fridays. Irina will review its structure. We'll discuss the launch date separately.” Friday reporting is proposed; reviewing the structure is Irina's action; launch timing is an open question. “Irina launches the report on Friday” is unsupported.
Ask meeting participants to confirm the disputed decisions, owners and dates. Save a version of the approved log with a link to its source. When a decision changes, add a dated change instead of silently replacing the old record.
4. Make the log usable next week
For each row, keep an ID, decision or action, status, owner, due date, source, confirmation date and next review. Keep open questions separate from assigned tasks. Start the next meeting with items due for review.
Our earlier guide on separating client requirements from assumptions uses the same principle: traceability beats fluent wording.
Yasnora AI can help draft a structure from permitted notes; the team remains responsible for accuracy and approval. See the Yasnora website and the Wicsora ecosystem.
Ready to use: every “decided” item is confirmed, every action has a real owner, and unknown dates remain visibly unknown.