qvib.pro
RU

Meeting summary: decisions and tasks

What for: pull the decisions, the tasks with their owners and the open questions out of meeting notes or a raw call transcript.

бесплатно новичок

Стратегия «дать опорный текст и извлечь из него» — OpenAI prompt-engineering checked 2026-06-01

Updated: 02.07.2026

$ You are an assistant who writes crisp meeting summaries. Work ONLY from the t…
Meeting summary: decisions and tasks

When to use it

After a call or a meeting, when you have a transcript or messy notes but need a clear outcome: what was agreed and who does what. The role is an assistant. Result: a 3-line summary + decisions + a task table with owners + open questions.

The prompt (copy and paste)

You are an assistant who writes crisp meeting summaries. Work ONLY from the text below, do not invent anything.
NOTES/TRANSCRIPT: "<PASTE>".

Deliver:
1. A 3-line summary — what was agreed.
2. Decisions — a list of the decisions made.
3. Tasks — a table: Task · Owner · Due date (if named).
4. Open questions — what was not settled / needs clarification.

If an owner or a due date is not named in the text — write "not specified", do not make it up.

Filled-in example

In <PASTE> — the transcript of an hour-long product call (from Read.ai/Whisper).

Expected AI answer: Summary (3 lines) — the team decided to ship feature X in beta, pushed the release by a week, onboarding design is needed. Decisions — a list. Tasks: "Build the onboarding UI · Anna · 15.06", "Prepare the beta list · Igor · not specified". Open questions — "plan pricing not agreed", "nobody assigned to beta support".

Variations

  • Follow-up email. "Format the summary as an email to attendees with a 'what I need from you' section."
  • My tasks only. "Pull out the tasks where the owner is ."
  • A series of meetings. Feed it several transcripts: "what from last time still has not been done."

Pro tips

  • "Work ONLY from the text" + "say not specified instead of guessing" is the core of this prompt: otherwise the AI will assign people tasks and deadlines that were never mentioned, and the follow-up will destroy trust.
  • Summary quality equals transcript quality: a clean transcript with speaker names gives accurate owners. A messy one gives confusion.
  • Meeting recordings with colleagues' voices are personal data: record with the participants' consent and do not pour sensitive negotiations into a public AI (Russian data-protection law 152-FZ).

Читать по-русски →