This guide gives you a reusable starting pattern. It is designed to help you see the work more clearly; it is not a substitute for judgment, source checking, or responsibility for the result.
Prepare the inputs before you ask for output.
The model only sees what you give it. Spend a few minutes naming the reader, desired result, and uncertain information. This makes a first draft easier to assess and reduces the need for decorative rewriting later.
Give the task a useful brief.
Prepare a working agenda for the next [meeting name] from the confirmed record below. Last meeting: [date and participants] Confirmed decisions: [list] Open questions: [list] Actions and stated owners: [list] New context since the meeting: [notes] Return: (1) meeting purpose, (2) the 3–5 items that need discussion or decision, (3) what participants should review beforehand, (4) the owner or input needed for each item, and (5) items that do not belong on this agenda. Do not reopen a confirmed decision unless the new context explicitly requires it. Do not assign an owner, date, or outcome that is not stated.
Replace every bracketed field with your real context. Read the output before reuse.
Four steps that keep the result usable.
- 1
Start from a confirmed meeting record rather than a raw transcript.
- 2
Separate decisions that are settled from questions that still need work.
- 3
Name the person or input needed for each agenda item when that information is known.
- 4
Have the meeting owner confirm the agenda before sending it.
An agenda is a decision surface
A long list of topics tells people where to look. A strong agenda tells them what needs to move. Give each item a purpose: decide, unblock, align, or prepare. The distinction makes it easier to protect the meeting from status updates that can happen asynchronously.
Carry forward uncertainty, not every sentence
Meeting notes often contain details that are useful only as context. Keep the question, dependency, or missing input that prevents progress, then link back to the record if more background is needed. That gives participants enough to prepare without turning the agenda into a transcript.
Do not turn preparation into a new commitment
AI can suggest a neat sequence, but it cannot know who has authority or capacity. If an owner, deadline, or decision status is not in the record, leave it visibly unresolved for the meeting owner to confirm.
A worked example: a monthly planning meeting
From last month’s record, the agenda might have three items: confirm the Q3 scope, decide who owns the onboarding checklist, and review the blocked hiring request. Each gets a purpose — decide, unblock, align — and a note about what to read beforehand. Items that were settled, like the approved budget, are left off so the meeting protects its time.
Protect preparation time
If the agenda requires reading a full transcript to prepare, it has not done its job. Keep the link to the record for detail, but make the pre-reading short and specific. An agenda that can be reviewed in ten minutes is more likely to be read than one that reopens everything.
Run a human check.
- Does every agenda item have a reason to exist now?
- Are settled decisions excluded unless new evidence requires review?
- Are missing owners or inputs visible rather than guessed?
- Could a participant prepare from the agenda without reading a full transcript?
AI is strongest here when it makes missing information, structure, and options easier to see. The moment an output becomes a claim, commitment, or decision, bring a person back into the loop.