Skip to main content

AI prompts for meetings and weekly planning

Agendas that defend their time-box. Weekly plans that name what gets cut. Status updates that fit on a screen. Built for the operational work that compounds quietly over time.

10 free prompts · tested in ChatGPT, Claude & Gemini · browse the full library

Generate a meeting agenda from a goal

Turn a vague 'we need to meet about X' into a real agenda.

Build a meeting agenda for this outcome.

Goal of the meeting: goal
Attendees: attendees
Length: minutes minutes

Produce:
1. **One-sentence purpose** — what success looks like by the end.
2. **Pre-read** (if any) — what each attendee should read first.
3. **Agenda** — 3–5 items, each with: topic, time-box (minutes), owner, expected output.
4. **Decisions required** — bulleted list of decisions that must be made (not discussions had).
5. **What this meeting is NOT for** — to defend the time-box.

Total time across agenda items must equal minutes minutes.
Why it works
The 'what this is NOT for' line stops scope-creep mid-meeting. The decision list prevents discussion-only meetings.
What to swap
Be honest about minutes — agendas built for 60 minutes don't compress to 30.

Weekly plan from a list of priorities

Turn a brain-dump of things to do into a realistic week.

I have these priorities for the week:
priority list

My constraints:
- I have roughly focus hours hours of focused work time across the week
- Recurring commitments: commitments
- Energy peaks: peak time

Output a Monday–Friday plan:
1. Schedule the top priorities into focus blocks during my peak times.
2. Be explicit about what does NOT make the cut — list it under "Cut this week".
3. For each day, give one sentence describing the day's single most important outcome.
4. If the priorities exceed the focus hours available, say so and recommend which to defer.
Why it works
Forces a cut list. Most planning fails because nothing is removed; this prompt makes deletion mandatory.
What to swap
Be specific in priority_list — 'do marketing' produces useless output; 'finalize landing page copy and send to designer' is workable.

Write a project status update from rough notes

Turn a handful of bullets into a structured update your team or stakeholders will actually read.

Write a project status update from the notes below.

Project: project name
Audience: audience (e.g. "eng team" or "exec sponsors")
Raw notes: notes

Output exactly this structure, no more:

**Status:** [On Track / At Risk / Blocked]
**Last week:** 2–3 bullets, outcomes not activities
**Next week:** 2–3 bullets, commitments not intentions
**Blockers / asks:** [specific ask + who can unblock it, or "None"]

Hard rules:
- ≤ 150 words total
- No "we made progress on", "we continued to", or "we worked on"
- Each bullet states what changed or what will change. Activity for its own sake does not belong here
Why it works
The word cap and the ban on activity-language are the two levers. 'We continued to migrate the database' is not a status; 'Migration is 60% done, on track for Friday' is.
What to swap
Include blockers even if you think they're minor — the reader can't unblock what they don't know about.

Turn a meeting transcript into a decision log

Capture what was agreed, not a summary of what was said.

Extract a decision log from the meeting transcript below.

Context: meeting context (e.g. "weekly product sync, 5 people, 45 min")

Transcript:
"""
transcript
"""

Produce:
1. **Decisions made** — each with: the decision in one sentence, who owns it, and any stated deadline.
2. **Assumptions embedded** — beliefs the team acted on without explicitly agreeing they were true.
3. **Open questions** — items discussed but not resolved, with who is responsible for resolving each.
4. **Next review trigger** — what event or date should prompt a revisit of these decisions.

Do not summarize the discussion. If something was talked about but not decided, it goes in Open questions. Decisions is only for what was actually decided.
Why it works
'Summarize this meeting' produces a recap of who said what. This prompt forces decision accounting — the only output that matters for async teams.
What to swap
Put the raw transcript in without cleaning it up — trimming it first removes the signal that separates firm decisions from soft ones.

Decide whether a meeting needs to happen at all

Test a proposed meeting against what it would actually produce before you send the invite.

I am about to schedule a meeting. Test whether it should exist.

Proposed meeting: meeting purpose
Who I would invite: attendees
Length: duration
What I want to walk out with: desired outcome

Answer in this order:
1. Is the desired outcome a decision, alignment, creation, or information transfer? Name which.
2. If it is information transfer, write the document or message that replaces the meeting, and stop.
3. If it is a decision: who actually decides? List anyone invited who is neither the decider nor an input to the decision, and say what they are there for.
4. What must be true before this meeting for it to reach its outcome? If any of it is missing, this meeting will fail — say so and name what to do first.
5. Only if it survives all of the above: write the invite, with the outcome stated in the first line and the pre-reading named.

Be blunt. Most proposed meetings should become a document.
Why it works
Sorting the purpose into four types first is what exposes the information-transfer meetings, which are the bulk of the waste.
What to swap
State `desired_outcome` as something you could hold in your hand afterwards. 'Discuss X' will always survive the test and should not.

Triage a backlogged inbox into actions

Turn a pile of unread messages into a short list of what actually needs you.

Triage the messages below. I have available time to deal with them.

My current priorities: priorities
What I am accountable for: my remit

Sort every message into exactly one bucket:
- REPLY NOW — blocked on me, and the block is real
- REPLY SHORT — needs a two-line answer; draft it
- DELEGATE — belongs to someone else; name who and draft the handoff line
- SCHEDULE — real work that needs a block of time; estimate how much
- ARCHIVE — no action, and say why in five words

Rules:
- Nothing goes in REPLY NOW because the sender is senior. Only because it is genuinely blocking.
- If REPLY NOW exceeds my available time, say which items slip and what the consequence is.
- Flag anything where the real deadline is earlier than the stated one.

Messages:
"""
paste messages here
"""
Why it works
Forcing exactly one bucket per message, with a time budget that can be exceeded, surfaces the triage decision instead of hiding it.
What to swap
Be specific in `my_remit`. It is the only thing that lets the model distinguish 'urgent' from 'urgent for someone else'.

Hand work over without it coming back

Write a delegation brief complete enough that the work does not bounce back to you.

Help me hand this work to someone else properly.

The work: the work
Who is taking it: their experience
Deadline: deadline
What "done" looks like: definition of done
Decisions they can make alone: their authority
Decisions they must bring to me: my authority

Write a brief that covers:
1. The outcome, and why it matters — the context I would otherwise keep in my head
2. Done criteria, stated so we could not disagree about whether they were met
3. The decision boundary, explicit in both directions
4. Known traps: what I would get wrong if I were doing this for the first time
5. Check-in points — the fewest that let me catch a wrong turn early
6. What I am deliberately not specifying, so they know where their judgment is wanted

Do not include instructions for how to do the work unless the method is itself a constraint.
Why it works
Most delegation fails at the decision boundary. Stating it in both directions is what stops the work bouncing back for approval.
What to swap
Point 6 is the one people skip. If you leave it out, you will get your own approach done slightly worse.

Audit a recurring meeting that has outlived its purpose

Work out whether a standing meeting should be kept, shortened, or killed.

Audit this recurring meeting.

The meeting: meeting name
Frequency and length: cadence
Who attends: attendees
Why it was originally set up: original purpose
What actually happens in it now: what happens now
The last three things it produced: recent outputs

Assess:
1. Total cost per month: attendee-hours, plus a note on the cost of the interruption around it
2. Does what happens now still serve the original purpose? If not, what purpose does it serve?
3. Which attendees could stop coming with no loss? Say what each currently contributes.
4. Could the cadence halve without anything breaking? What would break first?
5. Recommend one of: keep as is, shorten, reduce attendees, reduce frequency, replace with an async update, kill.
6. If your recommendation is anything other than keep, write the message announcing the change.

Commit to one recommendation. Do not suggest trialling several.
Why it works
Costing the meeting in attendee-hours changes the conversation from preference to arithmetic.
What to swap
`recent_outputs` is the load-bearing input. If you cannot name three, you already have your answer.

Review your week honestly

Turn a week of activity into a decision about what to do differently.

Review my week and tell me what to change.

What I planned to do: planned
What I actually did: actual
What I am accountable for this quarter: quarterly goals
Where the time went that I did not plan for: unplanned

Produce:
1. The gap between planned and actual, described as a pattern. Do not just list the misses
2. Which of the quarterly goals moved this week, and by how much. Name any that did not move at all.
3. The unplanned work: which of it was genuinely unavoidable, and which was me being the path of least resistance for someone else
4. The single change with the largest effect next week. Exactly one
5. What I should stop doing to make room for it, named specifically

Do not congratulate me on a busy week. Volume is not the metric.
Why it works
Asking for one change is what makes the review actionable. Five changes means none of them happen.
What to swap
`unplanned` is where the diagnosis lives. Be specific about who generated it as well as what it was.

Push back on scope without saying no flatly

Answer a new request by making the trade-off visible instead of absorbing it.

Someone has asked me to take on something new. Help me answer with the trade-off made visible.

The new request: new request
Who is asking and their authority over my priorities: requester
What I am currently committed to: current commitments
Realistic cost of the new request: estimated cost
What I would have to delay to do it: what slips

Write a reply that:
1. Takes the request seriously — no reflexive defense of my current list
2. States the cost concretely, in time and in what slips
3. Offers a real choice: do it and delay X, do a smaller version by Y, or do it after Z
4. Names who has to agree to the trade-off if it is not mine to make
5. Ends with a specific question, so the reply has to be a decision

Do not say yes and hope. Do not say no without a trade-off. Do not use "bandwidth", "capacity" or "prioritize" as a euphemism for no.
Why it works
Converting a yes/no into a visible trade-off moves the decision to whoever owns the priorities, which is usually not you.
What to swap
`what_slips` must be something the requester cares about. A trade-off against work they have never heard of is not a trade-off.

Frequently asked questions

Agendas that defend their time-box. Weekly plans that name what gets cut. Status updates that fit on a screen. Built for the operational work that compounds quietly over time. This page collects 10 free, copy-pasteable productivity & meetings prompts, each tested in ChatGPT, Claude, and Gemini with a one-line rationale and the fields you need to swap in.