Meeting Prep & Follow-ups with AI
AI earns its value in meetings before and after — not during. Pre-meeting research and agenda preparation take less than 10 minutes with AI. Post-meeting action item extraction from raw notes is instant. Both save real time. Using AI during a meeting to look things up or generate responses adds cognitive load and signals disengagement to the people in the room.
Pre-Meeting Research
Use AI to brief yourself before meetings with new clients, stakeholders, or domains you are less familiar with. Provide who is attending and what you are trying to achieve.
Prepare me for a meeting with [DESCRIBE WHO — e.g., "the VP of Engineering at a Series B fintech startup called Payro"].
Meeting goal: [WHAT YOU WANT TO ACHIEVE — e.g., "discuss integrating our data API into their payment reconciliation workflow"]
Provide:
1. Key context about their role and likely priorities (based on the company stage and function)
2. What they will probably care most about in this meeting
3. Five questions I should be ready to answer
4. Five questions I should ask to understand their situation
5. One thing I should not assume without confirming
Item 5 is the most valuable — it flags the assumption most likely to derail the meeting if left unchecked.
Agenda Generation
Provide the goal, attendees, and total time. Without attendee context, AI generates a generic agenda that serves no specific group.
Write an agenda for a [DURATION] meeting.
Goal: [SINGLE SENTENCE — what decision or outcome this meeting produces]
Attendees: [LIST ROLES — e.g., "product manager, 2 engineers, designer, customer success"]
Topics to cover: [YOUR LIST — can be rough bullet points]
What must NOT be on the agenda (to defer): [TOPICS TO EXCLUDE]
For each agenda item:
- Time allocation (minutes)
- Format: discussion / decision / update / brainstorm
- Who leads this item
- What the output of this item is (a decision, a shared understanding, a next step)
Also identify: which items could be async instead of in-meeting time.
The "what must NOT be on the agenda" field prevents scope creep — AI will fill time if you do not constrain it.
During the Meeting: What AI Cannot Do
Useful during meetings
- Timer / time-keeping (not AI — any clock)
- Quick factual lookup you would normally Google (fast, low-stakes)
- Live transcript tools (Otter.ai, Fireflies, Zoom AI Notes) that capture what is said
Counterproductive during meetings
- Generating responses to what someone just said — you are no longer listening
- Writing the agenda mid-meeting — should have been done before
- Research on topics raised unexpectedly — note it, research after
- AI note-taking while also participating — divided attention degrades both
Follow-Up: Notes to Action Items
This is where AI saves the most time. Paste rough meeting notes — incomplete sentences, scattered observations, speaker names — and AI produces a structured follow-up.
Convert these meeting notes into a structured follow-up summary.
Meeting: [NAME / DATE]
Attendees: [LIST]
Raw notes:
[PASTE YOUR NOTES — messy is fine]
Output format:
## Decisions Made
- [Each decision: what was decided and by whom]
## Action Items
- [Each action: OWNER — task — deadline (if mentioned)]
## Open Questions
- [Questions raised but not resolved — who will answer and by when]
## Next Meeting
- [Date / format / goal if mentioned]
Rules: only include information from the notes. Do not invent owners, deadlines, or decisions that are not in the notes.
The "do not invent" rule is critical — AI will assign owners and deadlines if you do not forbid it. Review the output against your notes before sending.
Async Meeting Alternatives
Many status updates, approvals, and FYI meetings do not need synchronous time. AI can draft the async alternative.
I have been asked to schedule a [DURATION] meeting for: [DESCRIBE THE PURPOSE]
Draft an async alternative that achieves the same goal without a synchronous meeting.
Format: a Loom/document brief that contains:
1. Context: what decision or update needs to happen and why it matters
2. The information attendees need (what I would have presented)
3. The specific input I need from each person (be explicit — vague "feedback" requests do not get responses)
4. Deadline: when I need their response
5. What happens if there is no response by the deadline
My raw notes for this meeting: [PASTE WHAT YOU WOULD HAVE COVERED]
The "what happens if no response" field is important — async alternatives fail when there is no consequence for non-participation.
Common Failure Modes
| Failure mode | Fix |
|---|---|
| Over-preparing: 30 questions for a 30-minute meeting | Limit to 5 questions; pick the two most important before the meeting starts |
| Generic agenda that ignores actual attendees | Always include attendee roles in the prompt; review agenda for fit before sending |
| AI-invented owners in action items | Add "do not assign owners not mentioned in the notes" to the prompt |
| Action items without deadlines | Add deadlines yourself after AI generates the list — AI cannot know your team's velocity |
| Async brief that nobody responds to | Make specific asks per person, not "please review"; state the consequence of non-response |
Checklist: Do You Understand This?
- What is the most valuable pre-meeting AI output — and what field in the prompt produces it?
- What does adding "what must NOT be on the agenda" prevent in agenda generation?
- Name two things that are counterproductive to do with AI during an active meeting.
- What rule must you add to the follow-up summary prompt to prevent AI from inventing action item owners?
- Write a pre-meeting research prompt for a call with a new engineer joining your team for an onboarding session.
- What makes an async meeting brief actually get responses — and what field in the prompt ensures this?