The line that decides whether anyone arrives prepared is the ask. Not the topic β the ask. "Review the Q3 plan" invites someone to turn up having skimmed it. "Decide: do we ship X in Q3 or Q4, and I need a yes or no today" tells them exactly what to walk in ready to do.
Rewrite every agenda item as a decision, and roughly a third of them stop needing a meeting at all. That is not a bug. Those items become messages, which is what they were.
The six formats
| Meeting | Length | Items | Output |
|---|---|---|---|
| Weekly team sync | 25 min | 3 | Blocked/unblocked list |
| 1:1 | 30 min | 3 | Two agreed next steps |
| Client call | 45 min | 4 | Decisions and owners |
| Decision meeting | 30 min | 1 | One recorded decision |
| Retro | 45 min | 4 | Two process changes |
| Board / stakeholder update | 60 min | 5 | Approvals and escalations |
1. Weekly team sync
Use when: the team needs to surface what is stuck, not report what was done.
Before (async, 5 min). Everyone posts: one thing finished, one thing blocked, one thing next.
In the meeting:
- Blocks (15 min) β each item either gets an owner and a date, or gets escalated. Nothing is discussed for more than four minutes; anything longer becomes its own meeting.
- Upcoming decisions (5 min) β anything landing this week that needs someone's attention before Friday.
- Announce and move on (5 min) β things people need to know, with no discussion attached.
The before step is the one teams skip, and it is the one that saves the time. Written updates are read in ninety seconds each; spoken updates take six minutes each and most of it is redundant.
2. 1:1
Use when: regular managerβreport meeting. Note that the agenda works better when the report sets it.
- Their item (10 min) β whatever they put on it. Do not reorder.
- Your item (10 min) β one thing, not three.
- Next two steps, said aloud (5 min) β each side states what they will do by when, before leaving.
The last step is what turns a pleasant conversation into something that changed anything. It takes two minutes and it is the whole point.
3. Client call
Use when: a scheduled check-in with an external party.
- Where things stand (5 min) β stated in their terms, not yours.
- Open decisions (20 min) β the ones only they can make, listed with what happens if nothing is decided.
- What we need from you (10 min) β named, dated, and assigned.
- Confirm next contact (2 min) β a date in both calendars before anyone leaves the call.
Send this agenda with the meeting request, not the day before. Clients who see it early arrive with answers; clients who see it an hour ahead arrive with questions.
4. Decision meeting
Use when: something has been discussed twice and still has not been decided.
Agenda, complete:
Decision: [one sentence, with the two or three real options]
Owner: [person whose call it is]
Background: [link β read before attending]
Time box: 30 minutes, ending with a recorded decision or a named date for one
Three rules. Background is read, not presented β anyone who arrives unprepared does not get a summary at everyone else's expense. One decision per meeting. And the outcome is written down while people are still in the room, including who disagreed and why, because "we decided" plus six different recollections is worse than no decision.
5. Retro
Use when: a project or cycle closes.
- What happened (5 min) β timeline, no interpretation.
- What worked (10 min) β debrief it while it is fresh, otherwise it becomes folklore.
- What did not (20 min) β addressed as process, not people. The question is what would have caught it earlier.
- Two changes (10 min) β two, owned, dated. Not six.
Two is the number for a reason. Teams that leave a retro with six action items complete none of them; teams that leave with two usually complete both.
6. Board / stakeholder update
Use when: reporting to people who are not in the work day to day.
- Where we are against plan (10 min) β including where you are behind, stated first.
- One thing they need to decide (20 min) β one, prepared, with your recommendation attached.
- Risks (15 min) β with probability and impact stated rather than adjectives.
- What we are not doing (10 min) β the explicit non-goals. This is the section that prevents surprises.
- Approvals needed (5 min) β each with a date attached.
Where AI helps, and where it costs you
Structuring a set of notes into agenda items is fast and works well. The risk is that a generated agenda tends to make every item look like a discussion topic, which is exactly the failure mode above. Read each line and add the ask: decide, approve, unblock, or just know. Lines with no ask get removed β which, in our testing, tends to shorten meetings more than any facilitation technique.
Editing time, from our 30-day benchmark
First draft from notes: under a minute. Converting topics into asks and deleting the rest: 3β5 minutes. If editing an agenda takes longer than the meeting will take, the meeting is probably too big.
Frequently Asked Questions
How far ahead should an agenda be sent?
With the invitation for anything involving an external party, and at least a day ahead for internal recurring meetings. Anything later and the prepared-reading step fails, which is the step that saves the time.
What if there is nothing to discuss?
Cancel it. A weekly meeting that survives on habit rather than content costs more in attention than it returns, and attendees who notice will start declining for the wrong reasons.
Should every item have a time box?
Yes for recurring meetings. Time boxes turn "we discussed X for forty minutes" from an accident into a decision someone can interrupt, and they make it socially easy to say take it offline.
Can AI draft agendas from my notes?
Yes, and it is quick. Add the ask to each item afterwards β decide, approve, or simply know β and delete anything that has none. That is where the time saving actually comes from.
Who should set a 1:1 agenda?
The report, where possible. When the manager sets it, the meeting drifts toward status; when the report sets it, it drifts toward what is actually on their mind, which is harder to get at and more valuable.