In many workplaces, meetings are one of the most time-consuming parts of the job. You build an agenda in prep, take notes on the day, write up the minutes afterward, share them with the people involved, and then chase the tasks that were decided — and this whole flow is actually packed with "busywork" you can hand to AI. In this chapter, step by step, you'll build the skill of lightening the entire meeting workflow with AI, from prep to cleanup. The "how to ask" you learned in the previous chapter on email and chat carries over directly.
The goal: let AI do the prep for the work before and after a meeting
The big picture: hand the meeting "busywork" to AI
Thinking of the meeting workflow as "one big task" makes it feel heavy, but break it into stages and each one has a place for AI. Hold the big picture first and it's obvious where to call in AI. Keep the "judgment, speaking, and agreement" that humans should do, and shift the "busywork" of prep and write-up around it to AI — that's the idea of this chapter.
Have AI help draft the agenda, organize the issues, and surface likely questions.
Automatically turn the audio of a recording or online meeting into text with AI.
Extract decisions and to-dos from the transcript and condense the key points.
Write notice messages and task lists for stakeholders to drive the next actions.
💡 You don't have to do it all at once. Start with the most painful stage. If writing minutes is a burden, do just "transcript → minutes"; if prep takes forever, do just "agenda creation." The relief of one thing getting easier is what pushes you to delegate the next stage.
Meeting prep — agendas and organizing the issues
Good meetings are decided in the prep. That said, building an agenda from scratch every time is a chore. Here, if you have AI produce a "first draft," instead of facing a blank page you can start by editing the version it gives you. The trick is to convey the purpose and theme, the participants, and the time you have.
Tell it "what this meeting needs to decide" and it proposes topics, time allocation, and goals — a blueprint for the meeting.
Hand over the theme and it lays out and organizes the issues, the angles likely to clash, and perspectives that tend to get missed.
Hand over your proposal and you can prepare in advance for the tough questions and concerns participants are likely to raise.
For example, to prep a meeting where you propose a new initiative, a single line like this is plenty.
Next week I'm running a 60-minute meeting with 6 internal stakeholders to decide on "introducing a new inquiry-handling workflow." The purpose is to decide whether to adopt it, plus owners and schedule.
Please create:
① a proposed agenda with time allocations
② 5 concerns or objections likely to come up in the room
③ hints on how to respond to each.
Take what fits your workplace's reality from what it produces. Tune it through conversation with "shorter" or "put the decision in the first half," and you'll have a running script for the day in one go.
Transcription — turn recordings and online meetings into text
The most painful part of making minutes is writing down what people say during the meeting. These days, "transcription" that automatically converts speech to text has reached a practical level, so you can hand almost all of this work to AI. There are broadly three methods. All follow the same flow: "recording (or audio file) → text."
Record an in-person meeting on a voice recorder or phone, then run the audio file through a transcription tool. The most versatile method.
Save it with your video-meeting tool's recording feature, then transcribe the audio or video. Many tools have built-in auto-captions and transcription.
A dedicated service where an "AI note-taker" joins the meeting and turns speech into text as it happens. The text is in your hands the moment it ends.
There are plenty of tools, and more appear every day. You don't need to fixate on one specific procedure. Choosing by "does the video tool I use have transcription?" or "is it a service I can hand my existing recording to?" is enough. Some distinguish between speakers, some are strong with multiple languages, and features vary. How to choose, and the representative tools, are collected in the article below.
📎 If you want the steps in detail, how to automate meeting minutes and transcription with AI and transcribing and captioning video and audio with AI are practical. Find the specific tool names and how to choose between them here.
⚠️ Recording and privacy come first. When you record a meeting, always get attendees' consent to record. Recording silently is a recipe for trouble. And if you're handing audio to an outside service, check whether it contains confidential or personal information, and follow your company's rules. For the standard on what's okay to enter, see information you should never enter into AI.
Auto-generating minutes — extract decisions and to-dos
Once you have the transcript, the next stage is turning it into readable minutes. A raw transcript is just a long conversation log, full of misstatements, tangents, and filler. This is where AI truly shines. Hand over the long text and ask it to "extract the decisions, to-dos, owners, and deadlines," and it shapes it into minutes with just the key points.
The key is to specify the items you want, concretely. Rather than just "make this into minutes," instructing the output format like below gives you a result you can use as-is.
Below is a transcript of a meeting. Create minutes from it.
■ Output format
・Meeting purpose (1 line)
・Decisions (bullet points)
・To-dos (item / owner / deadline in a table)
・Open items carried to next time
Cut the small talk and tangents and summarize only the facts. If the owner or deadline can't be read from what was said, write "TBD."
[Transcript]
(paste the transcript here)
The line "if the owner or deadline can't be read, write TBD" is what makes it work. By not letting AI fill in blanks on its own, you prevent fact and guesswork from getting mixed together.
⚠️ Always have a human check the minutes. AI can misread decisions or add deadlines nobody stated (hallucination). Check numbers, owners, deadlines, and amounts against the source before sharing. AI does the write-up, a human bears responsibility — don't break that division of labor.
Summarizing — turn a long record into three lines
Even with minutes in hand, not everyone reads them in full. For busy people, a "three-line summary you can grasp at a glance" is appreciated. AI is good at summarizing, but the trick is to specify "how long and how detailed." For the same minutes, a one-line summary for your boss and a more detailed bullet list for the person doing the work call for different summaries.
"Conclusion, next move, caution," one line each. For sharing instantly in chat.
5–7 items per issue. For people who couldn't attend to catch up.
Organize by specifying the lens, like "strengths / issues / on hold."
Summarize these minutes at two lengths.
① For a busy boss: 3 lines. Conclusion, next action, and concerns, one line each, no jargon.
② For the people doing the work: bullet points, up to 7 items. Make clear who does what by when.
[Minutes]
(paste the minutes here)
Summaries are more valuable the longer the original minutes or document. Hand over a document of dozens of pages and ask "first give me the big picture in three lines," and the time you spend deciding whether to read it shrinks dramatically.
Follow-up — notice messages and task lists
A meeting's real work is "after it ends." It only becomes results once you've told the stakeholders what was decided and gotten the tasks moving. You can hand this cleanup to AI too, using the finished minutes as raw material. Just hand over the minutes and ask, "draft a share-out email based on this" or "reorganize this into a task list."
Have it write a thank-you and key-points note for attendees and a share-out for those who missed it. You can specify recipients and tone.
Have it reorganize the to-dos into a list with owners and deadlines. It's handy to output it in a format you can paste into a spreadsheet or task tool.
Based on these minutes, write a share-out email to send to both attendees and those who missed it.
・A 3-line summary at the top
・Decisions and to-dos (with owners and deadlines)
・The next meeting's schedule
In a polite but easy-to-read, concise style. Also propose a subject line.
This is continuous with the email skills from the previous chapter. For tuning the style or writing differently for different recipients, the tips from Chapter 2, "Writing, email & chat," apply directly.
💡 Make "prep → transcript → minutes → summary → sharing" one flow. Once your body learns this pattern, you can reuse the same prompts for every meeting. Save your go-to instructions in a note and just paste them each time, and the post-meeting cleanup gets done in minutes.
- Break a meeting into prep → transcript → minutes → summary → sharing and AI has a place at each stage. Judgment is human, busywork is AI.
- In prep, have AI produce agendas, issues, and likely questions as a first draft. The trick is to convey the purpose, participants, and time available.
- Transcription has three methods — recording / online-meeting recording / real-time note-taker. Don't fixate on one tool, and put consent to record and handling of confidential information first.
- Extract minutes by specifying items — "decisions, to-dos, owners, deadlines," and summarize by specifying length and granularity. A human always checks the numbers and owners.
- Finally, produce a share-out email and task list from the minutes to drive the next actions.
Once the meeting busywork is lighter, it's time to speed up the "documents you present" themselves. In the next chapter, Chapter 4, "Creating documents and slides," let's build the skill of everything from structuring to turning it into slides.