Meetings are where decisions get made — and where information goes to die. The average professional sits through roughly 12 hours of meetings per week, and within 24 hours of any given meeting, most attendees have forgotten more than half of what was discussed. By the time someone needs to remember what was decided about the Q3 pricing pilot, the notes are either missing, scattered across three Slack threads, or buried in a notebook nobody can find.

Good meeting notes fix this. They turn an hour of conversation into a reusable artifact: decisions captured, action items assigned, context preserved, and context-shareable with anyone who couldn't attend. The problem is that taking great notes while also participating in a meeting is nearly impossible. If you're typing, you're not listening. If you're listening, you're not typing. The result is that most people either take no notes or take notes so sparse they're useless a week later.

AI has finally made this problem solvable at a level that actually works. With a recording or transcript and a well-structured prompt, you can produce clean, decision-ready meeting notes in under a minute — notes that are more complete, more consistent, and more useful than anything you could have written by hand while half-distracted.

In this guide, you'll find a complete toolkit of AI prompts for meeting notes and summaries, covering every common scenario: standard team meetings, client meetings, brainstorming sessions, all-hands, and one-on-ones. Every prompt is copy-paste ready and adaptable to your team's format and tone.

Why AI-Assisted Meeting Notes Win

Before we get to the prompts, let's be clear about why this is worth doing at all.

Participation over transcription: When AI captures the notes, you can actually be present in the meeting instead of half-listening while typing.
Consistency across meetings: The same prompt produces the same structure every time. Anyone reading your notes knows exactly where to find decisions, action items, and open questions.
Shareable with absentees: A clean AI-generated summary can be sent to someone who missed the meeting and they'll understand what happened without a 30-minute catch-up call.
Searchable history: Over time, your notes become a searchable decision log. "What did we decide about the API redesign in March?" becomes a 10-second search instead of a week-long archaeology project.
Faster follow-ups: Action items are already extracted and assigned. Follow-up emails and Slack posts can be drafted from the notes in seconds.

The leverage is real. The only question is whether you prompt well enough to capture it. Generic "summarize this meeting" prompts produce generic summaries that nobody reads. The prompts in this guide are designed to produce notes busy people actually use.

The Anatomy of Great Meeting Notes

Good meeting notes aren't a transcript. They're a structured artifact with specific components. Here's the anatomy we'll be prompting for:

Meeting meta: Title, date, attendees, and the meeting's purpose in one sentence.
Key decisions: What was decided, with enough context that the decision makes sense three months later.
Discussion summary: The shape of the conversation — what options were considered, what tradeoffs were weighed, where there was disagreement.
Action items: Who is doing what by when, with the owner named explicitly.
Open questions: Things that were raised but not resolved — important because these are what the next meeting needs to address.
Parking lot: Off-topic items captured for follow-up elsewhere so they don't get lost but also don't derail this meeting.
Next steps: When the group reconvenes and what's expected before then.

Every prompt below is built to surface or structure one or more of these elements. Used together, they cover the full meeting lifecycle.

Stage 1: Setting Up Before the Meeting

Great meeting notes start before the meeting. AI can help you set an agenda that gives the conversation enough structure to summarize cleanly afterward.

Prompt: Generate a Focused Meeting Agenda

You are a meeting facilitator. I'm running a [MEETING TYPE — e.g., weekly team sync, project kickoff, retrospective, client check-in] meeting.

Meeting context:
- Purpose in one sentence: [PURPOSE]
- Expected duration: [DURATION]
- Attendees and roles: [LIST]
- Key topics I know we need to cover: [LIST TOPICS]
- Decisions we need to make: [LIST DECISIONS IF ANY]
- Updates that are informational only: [LIST IF ANY]

Generate a realistic timeboxed agenda that:
- Allocates time proportional to the importance of each item, not equally.
- Groups related topics together so we don't context-switch.
- Puts the most important decision early enough that we won't run out of time.
- Includes 5-10 minutes at the end for action item review.
- Calls out any topic that should probably be deferred to async because it doesn't need a live discussion.

Format the agenda as a table with columns: Time | Item | Type (Decision / Discussion / Update) | Owner. Keep it under 12 rows.

The key insight in this prompt is the explicit ask to flag items that should be async. Most agendas waste time on updates that could be a Slack post. This prompt pushes the model to be ruthless about that.

Prompt: Define What "Good Notes" Means for a Specific Meeting Type

I'm setting up a recurring [MEETING TYPE] meeting in my company. I want to define a notes template that every meeting of this type will use, so notes are consistent across sessions and useful to readers.

Generate a notes template specifically for [MEETING TYPE] with these constraints:
- Total length when filled in: under 500 words.
- Sections, in order, each with a one-line description of what belongs there and a placeholder for the content.
- Sections should be tight — no section should ever produce more than 80 words of content for this meeting type.
- Include an explicit "Decisions" section and an explicit "Action Items" section, even if the rest of the template varies.
- Include a one-line "Context for absentees" line at the top designed to be forwarded to someone who missed the meeting.

Make the template copy-pasteable so I can drop it into a recurring doc.

This prompt helps you standardize. Once each meeting type has its template, your downstream AI prompts get more reliable because the expected output format is stable.

Stage 2: Turning a Transcript Into Structured Notes

This is the core use case — taking a transcript (or even just rough notes) and turning it into clean, structured notes. Most AI tools can transcribe meetings now; the leverage comes from what you do with that transcript.

Prompt: The All-Purpose Meeting Notes Generator

You are a meeting notes specialist. I'm pasting the transcript of a [MEETING TYPE] meeting below. Turn it into structured notes that a busy colleague could read in under 2 minutes and fully understand what happened.

[TRANSCRIPT]

Generate notes in this exact structure:

## Meeting Meta
- Title: [derive a clear title]
- Date: [if mentioned, else "unknown"]
- Attendees: [list names and roles if identifiable]
- Purpose: [one sentence]

## Decisions
- [Decision 1, with one line of context about why it was made]
- [Decision 2, with context]
- [Continue for all decisions made in the meeting]

## Discussion Summary
[3-5 sentences capturing the shape of the conversation: what options were considered, what tradeoffs came up, where there was disagreement. Do NOT list every opinion. Capture the arc.]

## Action Items
- [Owner] — [Action] — [Due date if mentioned, else "no date set"]
- [List every action item, even small ones]

## Open Questions
- [Questions raised but not resolved in the meeting]

## Parking Lot
- [Off-topic items captured for later]

## Next Steps
- [When the group reconvenes and what's expected before then]

Rules:
- Do not invent decisions that weren't made. If something was discussed but not decided, it goes in Open Questions, not Decisions.
- Do not invent action items or owners. Only extract what was actually mentioned.
- Preserve the actual language used for decisions whenever possible — paraphrase minimally.
- If a section has no content, write "None" rather than inventing filler.

The last set of rules is the most important part of this prompt. The biggest risk with AI meeting notes is that the model will fabricate decisions or action items to fill space. Telling it explicitly to write "None" when a section is empty prevents that.

Prompt: Ultra-Short Executive Summary

You are an executive assistant. I'm pasting a meeting transcript. Generate a single-paragraph executive summary (under 80 words) that a busy executive could read in 15 seconds and know whether they need to read the full notes.

[TRANSCRIPT]

The summary must include:
- The single most important decision made (if any).
- The single most important action item (if any).
- One item that's still unresolved.

Format as one paragraph. No bullet points. No headers.

This is the summary you put at the top of the notes doc or paste into Slack. The constraint to one paragraph and 80 words forces the model to prioritize.

Prompt: Action-Item-Only Output for Slack

You are a project coordinator. I'm pasting a meeting transcript. Extract ONLY the action items — nothing else.

[TRANSCRIPT]

Format as a Slack-ready message:

🎯 Action items from today's [MEETING TYPE]:

1. @owner — action — due date
2. @owner — action — due date
...

If an action item doesn't have a clear owner, flag it with ⚠️ and write "owner unclear" instead of guessing. If a due date wasn't mentioned, write "no date set" — do not invent one.

End with a single line: "Full notes: [LINK PLACEHOLDER]"

Most meeting notes end with a Slack post of action items. This prompt gives you exactly that, ready to paste.

Stage 3: Meeting-Type-Specific Prompts

Different meeting types need different note structures. Here are prompts tuned for the most common scenarios.

Prompt: Client Meeting Notes

You are a client success manager. I'm pasting the transcript of a client meeting. Generate notes optimized for sharing internally with our team — not for sending to the client.

[TRANSCRIPT]

Structure:

## Client Snapshot
- Client: [name]
- Our attendees: [list]
- Their attendees: [list with roles if identifiable]
- Meeting purpose: [one sentence]

## Client Sentiment
[1-2 sentences on the client's overall mood — enthusiastic, frustrated, neutral, anxious? Cite specific language from the transcript to support your read.]

## What the Client Asked For
- [Request 1 — quoted if possible]
- [Request 2]
- [Continue for every explicit client request]

## What We Committed To
- [Commitment 1 — with owner and date if mentioned]
- [Commitment 2]

## Risks & Escalations
- [Any client concern, complaint, or risk signal that we should flag internally. Quote the client's words.]

## Follow-Up Actions
- [Action — owner — date]

## Internal Notes
[Anything that came up that's relevant to our account strategy but shouldn't go in a client-facing summary.]

Rules: Quote clients verbatim when they express sentiment, complaints, or specific requests. Do not soften their language. If the client raised a complaint, capture it exactly.

The explicit instruction not to soften client complaints is critical. Most AI assistants instinctively sand down negative language, which is dangerous when you need to know a client is unhappy.

Prompt: Retrospective Notes

You are an agile coach. I'm pasting the transcript of a retrospective. Generate structured retro notes.

[TRANSCRIPT]

Structure:

## Retro Meta
- Team: [if identifiable]
- Period covered: [if mentioned]
- Attendees: [list]

## What Went Well
- [Item 1 — short, specific]
- [Item 2]
[Pull concrete examples from the transcript. Vague items like "good collaboration" should be rewritten with the specific example from the conversation.]

## What Didn't Go Well
- [Item 1 — short, specific]
- [Same rule: anchor to specific examples from the transcript]

## What We'll Change Next Sprint
- [Change 1 — with owner if assigned]
- [Change 2]

## Action Items
- [Action — owner — due date]

## Themes
[1-3 sentences on what patterns emerged across the discussion. What was the meta-story of this retro?]

Rules:
- Don't sanitize "what didn't go well" — capture the actual concerns raised, even if uncomfortable.
- If the team named a root cause, capture it. If they danced around a root cause, note that the discussion stayed at the symptom level.

Retrospectives only work if the notes are honest. This prompt explicitly resists the tendency to soften uncomfortable feedback.

Prompt: One-on-One Notes

You are an HR-savvy note taker. I'm pasting the transcript (or notes) from a manager-employee one-on-one. Generate structured notes that the manager can save privately and reference later.

[TRANSCRIPT]

Structure:

## 1:1 Meta
- Manager: [name]
- Employee: [name]
- Date: [if mentioned]

## Employee's Updates and Wins
- [Bullet list of what the employee raised — projects, wins, blockers]

## Employee's Concerns or Asks
- [Bullet list of any frustration, blocker, career question, or ask. Quote verbatim when sensitive.]

## Manager Feedback Given
- [Bullet list of feedback the manager offered — both praise and constructive]

## Commitments
- [Manager commits to: ...]
- [Employee commits to: ...]

## Follow-Up for Next 1:1
- [Topic 1 to revisit]
- [Topic 2]

Rules:
- Keep the tone neutral and professional — these notes might be read months later.
- Capture sensitive topics (career dissatisfaction, interpersonal conflict, burnout signals) in the employee's own words when possible.
- Do not editorialize or add judgment.

One-on-one notes are a special case because they often involve sensitive topics. The explicit instruction to preserve the employee's own words matters here.

Prompt: Brainstorming Session Notes

You are a facilitator. I'm pasting the transcript of a brainstorming session. The goal of the session was: [GOAL OF BRAINSTORM].

[TRANSCRIPT]

Structure:

## Brainstorm Meta
- Goal: [restate the goal]
- Participants: [list]
- Duration: [if mentioned]

## Ideas Generated
Group ideas into themes. For each theme:
### [Theme name]
- [Idea 1 — one line]
- [Idea 2]
- [Idea 3]
[Continue for all themes. Don't lose any ideas — even ones that seemed half-baked in the moment.]

## Ideas That Got Traction
- [Ideas the group circled back to or built on]

## Ideas Flagged for Further Exploration
- [Ideas worth a deeper dive but not yet refined]

## Decisions Made
- [Specific decisions, if any — most brainstorms don't make decisions, so this might be "None"]

## Next Steps
- [Action — owner — date]

Rules:
- Capture every idea, even ones dismissed quickly. Brainstorm value is often in ideas that didn't seem important in the moment but do later.
- Don't editorialize on idea quality. Just capture them.

Brainstorms produce huge volumes of ideas and most notes lose half of them. This prompt explicitly forces the model to keep every idea, even discarded ones.

Stage 4: From Notes to Follow-Up

The notes aren't the end of the workflow. The leverage compounds when you turn notes into follow-ups, recaps, and searchable history.

Prompt: Draft a Meeting Recap Email

You are a professional communicator. I'm pasting structured meeting notes. Draft a recap email suitable for sending to attendees and key absentees.

[NOTES]

Requirements:
- Subject line: actionable and under 70 characters.
- Greeting: warm but not effusive.
- Body: 2-3 short paragraphs covering decisions and discussion shape.
- Action items as a clear bulleted list with bolded owner names.
- Closing: one sentence on what's expected before the next meeting.
- Tone: professional, concise, scannable.

Do not use the words "I hope this email finds you well" or any similar filler opener.

Prompt: Update a Decision Log

You are a project historian. I'm pasting meeting notes. Extract every decision and append it to an ongoing decision log in this exact format:

[NOTES]

For every decision, output a row in this format:

### [Date] — [Meeting title]
- **Decision**: [one sentence]
- **Context**: [one sentence on why this decision was made]
- **Decided by**: [names if identifiable]
- **Relevant to**: [project / workstream this decision affects]
- **Status**: Active (unless overturned later in the transcript)

If the meeting overturned an earlier decision, output a row for the new decision AND note which previous decision it overturned in the Context field.

Skip any "decision" that was actually just a discussion without resolution.

This builds a decision log over time — one of the most valuable long-term artifacts any team can maintain.

Prompt: Generate Follow-Up Questions for the Next Meeting

You are a meeting facilitator. I'm pasting notes from our most recent [MEETING TYPE] meeting. Generate a focused list of follow-up questions and topics that the next meeting needs to address.

[NOTES]

Structure the output as:
1. Open questions from the last meeting that still need answers.
2. Action items due before the next meeting that we should check on.
3. New topics that emerged from the last discussion but weren't resolved.
4. One or two questions a good facilitator would add to surface what we're not talking about yet.

Keep it under 10 items total. Prioritize ruthlessly.

That last prompt turns every meeting's notes into the seed for the next meeting's agenda. Once you start using it, your meetings naturally chain together instead of feeling like disconnected conversations.

Stage 5: Running With AI Assistance Live

If you want to take it further, you can have AI running alongside your meeting in real time. Most modern meeting tools (Otter, Fathom, Read.ai, and built-in options in Teams, Zoom, and Google Meet) handle transcription. The prompts in this guide work just as well against live transcripts as against recorded ones — just run them at the end of the meeting with the transcript and you'll have notes before you've even left the room (virtually).

For the highest leverage, drop your standard template (from Stage 1) into your AI tool's prompt configuration so every meeting of a given type produces notes in your house format automatically. Then the prompts in this guide become your refinement and follow-up layer rather than your starting point.

Common Mistakes to Avoid With AI Meeting Notes

AI meeting notes are powerful but easy to get wrong. Here's what to watch for:

Fabricated decisions: The most dangerous failure mode. AI fills in plausible-sounding decisions that were never actually made. Always review Decisions and Action Items against the transcript before sharing.
Generic summaries: A 5-bullet summary that could describe any meeting is useless. Force the model to extract specifics — names, numbers, quoted language.
Lost quiet voices: AI over-indexes on whoever talked most and misses the one person who made a critical point in 10 seconds. Skim the transcript yourself for moments where quieter voices surfaced something important.
Over-formatting: Adding 12 sections to a 20-minute meeting is overkill. Match the note structure to the meeting's actual weight.
No follow-up loop: Notes that nobody reads and nobody acts on are theater. Pair every notes prompt with a follow-up prompt (recap email, decision log, next agenda).
Privacy drift: Meeting transcripts often contain sensitive content. Make sure your AI tool's data handling is appropriate for what was discussed — not every conversation belongs in every model.

The Bottom Line

Good meeting notes are the difference between a team that learns from its own conversations and a team that keeps re-explaining the same context every week. AI makes consistent, high-quality notes achievable for every meeting, not just the high-stakes ones.

The prompts in this guide are starting points. Tailor the brackets, save your favorites as templates, and run them against whatever model fits your workflow — ChatGPT for general drafting, Claude for nuanced prose and sentiment, Gemini for fast in-Google-Workspace use. The structure matters more than the model.

If you want to take this further — turn these one-off prompts into a versioned, testable library your whole team can use consistently — sign up for PromptWright. We help operations and people teams build reusable prompt workflows so the prompts that work get shared, improved, and standardized instead of living in one person's private notes.

Frequently Asked Questions

Do I need a meeting recording to use AI for notes?

Not necessarily, but it helps a lot. AI works best from a transcript, and a transcript comes from a recording. If you can't record — client confidentiality, sensitive one-on-ones — you can still feed AI your rough handwritten notes and get structured output. The result will be less complete but still useful.

Which AI tools produce the best meeting notes?

For general-purpose prompts against transcripts, Claude produces the most natural prose and is strong at sentiment. ChatGPT is reliable for structured extraction. Gemini is fast and integrates well if you live in Google Workspace. Dedicated tools like Otter, Fathom, and Read.ai handle transcription and basic summarization end-to-end. Use dedicated tools for capture, then layer the prompts in this guide for customized output.

How do I prevent AI from inventing action items?

The most effective guardrail is to explicitly tell the model: "Only extract action items actually mentioned in the transcript. If no action items were assigned, write None." Then review the Action Items section against the transcript yourself before sharing. A 30-second verification catches most fabrications.

Should meeting notes be shared with everyone who attended?

Usually yes, with exceptions. Standard team meetings, retrospectives, and brainstorms benefit from wide sharing. Client meetings, one-on-ones, and sensitive discussions should be shared more selectively. The notes capture what was said, so think about whether what was said was meant for a wider audience.

How long should AI-generated meeting notes be?

For a 30-60 minute meeting, aim for 300-500 words of structured notes. Shorter is usually better — the goal isn't to capture every word, it's to capture what someone would need to know a month from now. If your notes are running over 600 words regularly, your prompt is probably extracting too much discussion and not enough decision.

AI Prompts for Meeting Notes and Summaries: Capture Every Decision Without Burning Out