The Real Cost of Your Meetings

Take your last week of recurring meetings. Multiply attendees by minutes by a conservative loaded hourly rate. The number is uncomfortable. Most of that spend buys very little, because the meeting is doing three jobs at once: surfacing information that could have been read, debating decisions nobody framed, and assigning work nobody captured.

AI doesn't fix bad meetings. It exposes them. Once a transcript, a summary, and an action-item list are free, the meeting itself has to justify the time. That is the right pressure to apply. Your job is to use that pressure to shrink the live portion to the part that genuinely needs humans in a room.

The target you should hold yourself to is concrete: a 60-minute standing meeting compressed to 25 minutes of live conversation, with a written trail that is more useful than the recording.

Pre-Reads That Replace the First 20 Minutes

The first third of most meetings is status theater — people narrating what they did so others know. Kill it. Have AI draft a pre-read from raw inputs (Slack threads, tickets, last week's notes, dashboard exports) and circulate it 24 hours before.

A prompt that works:

You are drafting a 1-page pre-read for a 30-minute weekly product
sync. Audience: 5 engineers and a designer who will read this in
under 4 minutes.

Inputs (raw, unedited):
- Linear export of tickets closed and opened this week: [paste]
- Slack #product-launch thread from Mon-Wed: [paste]
- Last week's meeting notes: [paste]

Output exactly three sections:
1. What shipped (5 bullets max, plain language, no jargon)
2. What's blocked, with the specific question the meeting needs to answer
3. Decisions needed today (numbered list, each phrased as a yes/no
   or pick-one choice)

Skip anything that is on track and not blocked. If a section is empty,
write "None this week" — do not invent content.

The third section is the unlock. When the meeting opens, you skip status and go straight to decision 1. If there are no decisions and nothing blocked, cancel the meeting and send the pre-read as the update. People will thank you the first time you do this.

A skeptical note: AI summaries hallucinate confidence. Always have one human (often you) skim the pre-read before it ships. The cost of a wrong "shipped" bullet circulating to the team is higher than the time saved.

Live Notes Without a Notetaker

If your tooling allows it and your org's policy permits it, run a transcription tool (Otter, Granola, Fireflies, Zoom's built-in, Google Meet's). Announce it at the top — "I'm recording for notes, transcript goes in the doc, tell me if you want something off the record." That sentence covers consent, sets the norm, and gives people an out.

What changes during the meeting: you stop typing. Your attention goes to facilitation — calling on quiet people, cutting tangents, closing decisions. The transcript handles memory.

Two habits make the transcript usable later:

• Say decisions out loud, in full sentences. "Decision: we ship the export feature behind a flag on Friday, Priya owns the rollout." The transcript will catch that exact phrasing, and AI will lift it cleanly into the summary.
• Name the owner and the date for every action. "Action: Marco writes the migration plan, draft to me by Wednesday EOD." Vague verbs like "look into" or "circle back" produce summaries you cannot follow up on.

Train yourself to interrupt and restate when a decision happens implicitly. It feels mildly awkward for two weeks, then it becomes how your team operates.

Turning the Transcript Into a Clean Trail

After the meeting, feed the transcript through a structured summarizer. Don't ask for "a summary" — that gives you a wall of paragraphs nobody reads. Ask for the artifact you actually need:

You are processing a meeting transcript into a structured note for
async readers who were not in the room.

Transcript: [paste]

Produce exactly this structure, in markdown:

## Decisions
- One bullet per decision. Format: "Decision: <what>. Owner: <name>.
  Context: <one sentence on why>."

## Actions
- One bullet per action item. Format: "<Owner> — <verb + specific
  deliverable> — due <date>."
- If no date was given, write "due: NEEDS DATE" so I can chase it.

## Open Questions
- Anything raised but not resolved. Include who needs to answer it.

## Risks Flagged
- Concerns or blockers mentioned. One line each.

Rules:
- Use names exactly as they appear in the transcript.
- Do not summarize discussion. Only capture decisions, actions,
  questions, and risks.
- If a section is empty, omit it.

Paste it into the team doc or channel within an hour. The "NEEDS DATE" trick is small but powerful — it surfaces every vague commitment so you can pin it down before people forget what they agreed to.

For sensitive meetings (performance, comp, legal, anything involving customer data) skip the cloud transcription. Take notes yourself, or use an on-device tool. The privacy cost is not worth the convenience. Chapter 11 goes deeper on what you can and cannot route through AI.

The 25-Minute Standing Meeting

Once pre-reads and AI summaries are running, restructure the meeting itself:

• Minutes 0-2: Confirm everyone read the pre-read. If not, give them two minutes to skim.
• Minutes 2-15: Walk the decisions list. Each decision gets a hard cap — three minutes of discussion, then a call. If you cannot decide in three minutes, you do not have enough information; assign someone to gather it and move on.
• Minutes 15-22: Blocked items. Whose help is needed, by when.
• Minutes 22-25: Read back the decisions and actions out loud. The transcript will catch this, and the summary will be cleaner because of it.

End early when you can. A meeting that ends at minute 18 is not a sign you ran it badly — it is a sign the system is working. The pre-read absorbed the status, the AI absorbed the note-taking, and the humans did the thing only humans can do: decide.

If meeting craft is something you want to build deeper, the AI for Managers Playbook course works through facilitation patterns and decision frameworks in more depth, and AI Writing & Content Creation sharpens the summaries and follow-ups you send afterward.