How to Prompt AI to Write Meeting Agendas

The fastest way to make a meeting run long is to hand it an agenda that mixes a status update, a debate, and a brainstorm with no indication of which is which. Knowing how to prompt AI to write meeting agendas comes down to one idea: tie the agenda's structure to the type of each item, info-share, decision, or brainstorm, because each type needs a different time box and a different kind of prep. A prompt that skips that distinction will hand you a tidy-looking list that still runs forty minutes over.

Info-share, decision, and brainstorm items are not the same item

This distinction follows the prompt engineering framework behind this: give the model a taxonomy before you give it content, and it stops treating every line item the same way. An info-share needs a named presenter and a hard time cap, because no group decision is happening. A decision item needs a named decision-owner and the exact question that must be answered before the meeting ends. A brainstorm needs a facilitator and a defined output, three shortlisted ideas, not the open-ended word “discuss.” Three different item types, three different sets of instructions.

The structured agenda prompt

This is the same structured approach applied to status reports: decide the categories first, then generate content inside them, instead of writing free-form and hoping structure emerges on its own. Here is the prompt, run against a real list of raw agenda items for a six-person marketing team's weekly meeting:

“Turn the raw agenda items below into a structured meeting agenda for a 40-minute meeting. For each item, classify it as info-share, decision, or brainstorm. Info-share items get a 5-minute cap and a named presenter. Decision items get a named decision-owner and the specific question that must be answered by the end of the item. Brainstorm items get a named facilitator and a defined output, not just 'discuss.' If an item has no owner listed, or does not actually require a decision or a defined output, do not schedule it. Flag it instead with a one-line suggestion for handling it outside the meeting. Raw items: Q3 campaign results, CRM vendor decision, holiday campaign ideas, Sarah's onboarding, open floor.”

Before and after: the same five items, two agendas

The raw list, turned into an agenda without this structure, looks like most agendas do:

1. Q3 campaign results

2. CRM vendor decision

3. Holiday campaign ideas

4. Sarah's onboarding

5. Open floor

Nothing here says which items need a decision, who owns that decision, or how long any of it should take, which is exactly why it runs long. Run the same five items through the structured prompt and you get this instead:

Open floor disappeared entirely, because it never had an owner or an output, which is the exact flag the prompt is built to catch.

Why the owner-and-decision flag is the part that matters

The single biggest cause of a meeting running long is not too many items, it is items with no owner and no decision attached sitting on the agenda anyway. An unassigned task is exactly the failure mode a handover document prompt with the same owner-and-decision discipline is built to catch. An agenda item works the same way: it either gets a decision-owner or it gets cut, there is no third option.

An AI meeting agenda template for recurring meetings

Save the structured prompt above as a reusable ai meeting agenda template, and swap in a fresh list of raw items each week; the classification logic does not change from meeting to meeting. It is the same fix as a SWOT analysis prompt built the same structured way: give the model categories before content, and make it flag whatever does not fit a category instead of quietly including it anyway.

How to prompt AI to write meeting agendas for different meeting types

The same structured agenda prompt adapts to different meeting types just by changing what you expect back, not by changing the classification logic itself. A daily standup is almost entirely info-share items with a 60 to 90 second cap each, so the decision-owner field goes unused most days. Sprint planning is close to the opposite: mostly decision items, each with a named owner and a specific question, is this ticket in scope for this sprint. A 1:1 rarely needs the facilitator field a brainstorm calls for, but it uses the no-owner flag the most, since 1:1s are exactly where vague, ownerless topics like “catch up on stuff” tend to survive the longest.

• Daily standup: almost entirely info-share, short caps, no decision-owner field needed

• Sprint planning: mostly decision items, each with a named owner and a specific question

• 1:1 check-in: mixed types, but the no-owner flag catches vague topics like “catch up on stuff”

• All-hands: pure info-share, multiple presenters, no brainstorm items at all

Frequently asked questions

What's a good prompt for a weekly meeting agenda that repeats every week?

Keep the classification instructions fixed, info-share, decision, brainstorm, plus the no-owner flag, and only change the raw item list each week. Treat the prompt itself as a template, not something you rewrite from scratch.

How do I stop meetings from running long even with a good agenda?

Enforce the time boxes the prompt assigns. A structured agenda only works if someone is willing to cut off an info-share at five minutes instead of letting it become a discussion.

What if an item genuinely needs a decision but also needs some brainstorming first?

Split it into two agenda items with two time slots. A single item that mixes brainstorming and deciding is how meetings drift past their scheduled end time.

Should every meeting agenda include a flagged-items section?

Only if the raw list actually contains something without an owner or a decision. An agenda with nothing to flag does not need an empty flagged-items header.

Does this work for meetings run over video calls with async prep?

Yes, paste the flagged items into the pre-read doc so people arrive with context on the decision items specifically, that is where a structured, AI-written agenda saves the most time.