Meeting Summary Agent
operations
The Problem
Meetings end and everyone remembers something different. Action items get lost, decisions aren't documented, and someone spends 30 minutes writing a summary nobody reads.
The Solution
An AI agent that listens to your meeting (or reads the transcript), extracts key points, decisions, and action items, then distributes a clean summary to all participants.
How It Works
- Capture: Meeting is recorded via Zoom, Teams, or Google Meet
- Transcribe: Audio converted to text (Whisper, Deepgram, or built-in)
- Analyze: Agent identifies speakers, topics, decisions, and tasks
- Structure: Creates formatted summary with sections
- Distribute: Sends to Slack, email, or Notion automatically
Tool Stack
- Zoom/Teams/Meet: Recording source
- Deepgram or Whisper: Transcription
- Claude/GPT-4: Summary generation
- n8n: Workflow automation
- Notion/Slack: Output destination
Example Prompt
Analyze this meeting transcript and provide a structured summary:
## Meeting Summary
- Date: [extract]
- Participants: [list names mentioned]
- Duration: [estimate from transcript]
## Key Discussion Points
[3-5 bullet points of main topics]
## Decisions Made
[List any explicit decisions with who made them]
## Action Items
| Task | Owner | Deadline |
|------|-------|----------|
[Extract specific commitments with names]
## Open Questions
[Any unresolved items needing follow-up]
Transcript:
{{transcript}}
Implementation Tips
- Always identify speakers if possible (improves action item assignment)
- Keep summaries under 500 words—brevity increases readership
- Send within 10 minutes of meeting end while context is fresh
- Include a link to full transcript for reference
When to Use This
- Team has 5+ recurring meetings per week
- Action items frequently slip through cracks
- New team members need to catch up on past discussions
- Compliance requires meeting documentation
Limitations
- Speaker identification isn't perfect in group settings
- Won't capture non-verbal cues or whiteboard content
- Confidential meetings need extra security consideration
- Quality depends on audio/transcript quality