How to Use NotebookLM for Meeting Notes: Automated Action Item Extraction and Follow-Up Tracking
The Meeting Notes Problem Nobody Has Solved
Every organization has the same meeting notes problem. Someone takes notes (or nobody does). The notes are incomplete, biased toward what the note-taker found interesting, and stored in a document that nobody reads again. Action items are buried in paragraphs of text. Decisions are recorded ambiguously. Three weeks later, the team disagrees about what was decided because nobody can find the relevant notes.
Meeting transcription tools (Otter.ai, Fireflies, Grain) partially solve this by capturing everything that was said. But they create a new problem: a 60-minute meeting produces 8,000-12,000 words of transcript. Nobody reads 10,000 words to find the 3 action items and 2 decisions that matter.
NotebookLM solves the extraction problem. Upload your meeting transcripts (one or many), and NotebookLM becomes a queryable knowledge base of everything that was discussed. You can ask specific questions: “What did we decide about the API redesign?” “Who is responsible for the Q2 budget proposal?” “What concerns were raised about the migration timeline?” — and get precise, sourced answers.
This guide covers the complete workflow from transcript upload to actionable meeting intelligence.
What You Need
- Google NotebookLM (free with a Google account)
- Meeting transcripts from any source:
- Zoom (auto-transcript or Otter.ai integration)
- Google Meet (built-in transcription)
- Microsoft Teams (built-in transcription)
- Otter.ai, Fireflies, or Grain exports
- Manual notes in Google Docs or text files
- A consistent workflow for uploading transcripts after each meeting
Step 1: Set Up Your Meeting Notebook
Notebook Organization
Create one notebook per recurring meeting series:
Notebooks: - "Weekly Team Standup" (contains all weekly standup transcripts) - "Product Planning" (contains all product planning transcripts) - "Leadership Sync" (contains all leadership meeting transcripts) - "Client: Acme Corp" (contains all meetings with this client) - "Project: Migration" (contains all meetings about the migration project)
Why separate notebooks? NotebookLM searches within a single notebook. If all your meetings are in one notebook, a query about “the budget” returns results from every meeting where budget was mentioned. Separate notebooks keep contexts clean.
When to combine: Put related meetings in the same notebook when you want cross-meeting intelligence. For example, all product planning meetings in one notebook lets you ask “How has the roadmap changed over the past 3 months?”
Uploading Transcripts
For each meeting:
- Export the transcript from your recording tool (TXT, Google Docs, or PDF)
- Upload to the appropriate NotebookLM notebook as a source
- Name the source clearly: “2026-03-24 Product Planning” (date + meeting name)
Transcript format tips:
- Include speaker labels if available (“Sarah: We should prioritize the API…”)
- Include timestamps if available (helps NotebookLM locate specific moments)
- Remove auto-generated filler (“um”, “uh”, “you know”) if your tool includes them — they add noise without value
Step 2: Extract Action Items
The Action Item Query
After uploading a transcript, ask NotebookLM:
"Extract all action items from this meeting. For each action item, provide: 1. Task description (what needs to be done) 2. Owner (who is responsible) 3. Deadline (if mentioned, or 'no deadline stated') 4. Context (why this task matters — 1 sentence) 5. Dependencies (does this depend on another task?) Format as a numbered list. Only include items where someone committed to doing something specific."
What Counts as an Action Item
NotebookLM should distinguish between:
Action items (include):
- “Sarah will send the updated proposal by Friday” — clear owner, clear task, clear deadline
- “Engineering needs to evaluate the three vendor options” — clear team, clear task, no deadline
- “Let’s set up a follow-up meeting next week to review progress” — clear action, implied owner
Not action items (exclude):
- “We should think about expanding to Europe” — aspirational, no commitment
- “That’s a good point about security” — acknowledgment, not a task
- “Maybe we could use a different approach” — suggestion, not a commitment
If NotebookLM includes vague items, refine:
"From the action items you identified, remove any that are aspirational or speculative. Only keep items where a specific person or team committed to a specific deliverable."
Handling Missing Owners or Deadlines
Many action items emerge without clear ownership:
"For action items without a clear owner, note who was speaking when the task was discussed — they are the likely default owner. For items without deadlines, flag them as 'needs deadline assignment' so they are addressed in the next meeting."
Step 3: Generate Audience-Specific Summaries
Executive Summary (2-3 minutes to read)
"Write a 200-word executive summary of this meeting for senior leadership. Include only: 1. Decisions made (with brief rationale) 2. Key risks or blockers identified 3. Resource or budget requests 4. Timeline changes Do not include discussion details or brainstorming — only outcomes and items requiring executive attention."
Team Lead Summary (5 minutes to read)
"Write a summary of this meeting for the team lead who was not present. Include: 1. All decisions made and the reasoning behind them 2. Action items assigned to their team members 3. Dependencies on or from other teams 4. Open questions that need their input 5. Any changes to previously agreed plans Keep it under 500 words."
Individual Contributor Summary (task-focused)
"Write a summary for [person's name] that includes only: 1. Action items assigned to them 2. Decisions that affect their current work 3. Information they need to know for their tasks 4. Questions that were asked about their area that they should be prepared to answer Keep it under 200 words — just what they need to act on."
Stakeholder Update (external)
"Write a client-facing summary of this internal planning meeting. Include only information appropriate for the client: 1. Timeline updates 2. Deliverable status 3. Any questions we need from them 4. Next meeting date Exclude: internal debates, resource constraints, budget discussions, and any internal-only concerns."
Step 4: Track Decisions Over Time
The Decision Log Query
With multiple meeting transcripts in the same notebook:
"Search all meeting transcripts and create a decision log. For each decision: 1. Date of the meeting 2. What was decided 3. Who made the decision (or who was present) 4. What alternatives were considered (if discussed) 5. Rationale for the decision 6. Any conditions or caveats Sort chronologically, most recent first."
Tracking Decision Changes
Over time, decisions get revisited and changed:
"Has the decision about [specific topic] changed since it was first made? If so: 1. What was the original decision? (date, context) 2. When was it revisited? (date, what triggered the review) 3. What changed? (new decision) 4. Why did it change? (new information, changed requirements)"
This is invaluable for understanding why things are the way they are. Six months from now, when someone asks “Why are we using vendor X instead of vendor Y?”, the answer is in the meeting notebook.
Commitment Tracking
"Find all commitments made by [person name] across all meetings in this notebook. For each commitment: 1. What they committed to 2. When they committed (meeting date) 3. Deadline (if any) 4. Was this commitment fulfilled? (check subsequent meetings for follow-up)"
Step 5: Identify Recurring Topics
Pattern Analysis
With 10+ meeting transcripts in a notebook:
"Analyze all meetings in this notebook and identify topics that have been discussed in 3 or more meetings. For each recurring topic: 1. How many meetings has it appeared in? 2. What was the discussion about each time? 3. Has progress been made between discussions? 4. Is there a pattern of discussing without resolving? Flag any topic that has appeared in 5+ meetings without a clear resolution — these are likely systemic issues."
The “Groundhog Day” Detector
Some teams repeatedly discuss the same issue without resolving it:
"Identify any issues that have been raised in multiple meetings where the same concerns are expressed each time but no action is taken. These are 'Groundhog Day' issues. For each: 1. What is the issue? 2. How many times has it come up? 3. What blocks resolution? 4. What would it take to resolve it permanently?"
This analysis often reveals organizational dysfunction — an issue that keeps coming up but never gets fixed usually indicates a missing decision-maker, misaligned incentives, or a resource constraint nobody wants to acknowledge.
Topic Evolution
"How has the discussion about [topic] evolved over the past 3 months? Show the progression: - First mention: what was the initial framing? - Subsequent discussions: how did understanding change? - Decisions made along the way - Current status: where does this topic stand now?"
Step 6: Generate Pre-Meeting Briefs
Before the Next Meeting
"Based on all previous meeting transcripts, generate a pre-meeting brief for the next [meeting name] session. Include: ## Open Action Items (Items assigned but not yet confirmed complete) ## Decisions Pending (Topics discussed but not decided — need resolution) ## Recurring Issues (Topics that keep coming up — need dedicated time) ## Follow-Ups Due (Commitments that should have been completed by now) ## Suggested Agenda (Based on open items and pending decisions, suggest an agenda for the next meeting)"
This pre-meeting brief ensures nothing falls through the cracks. It takes 2 minutes to generate and saves 10-15 minutes of the meeting that would otherwise be spent on “Where were we?” and “What did we decide last time?”
Advanced Workflows
Cross-Meeting Intelligence
When you need to connect information across meeting series:
Option 1: Create a dedicated notebook and upload transcripts from multiple meeting series that relate to a specific project or initiative.
Option 2: Query each notebook separately and synthesize manually.
Example cross-meeting query:
"What has been said about the API redesign project in this notebook? Include all mentions, decisions, concerns, timeline discussions, and resource commitments."
Onboarding New Team Members
Upload the last 3 months of team meeting transcripts to a notebook. A new team member can ask:
"I'm new to this team. Based on these meeting transcripts: 1. What are the team's current top priorities? 2. What major decisions have been made recently? 3. What ongoing debates or open questions exist? 4. Who seems to own which areas? 5. What jargon or project names should I know?"
This gives new hires context that would normally take weeks of attending meetings to acquire.
Meeting Effectiveness Analysis
"Analyze the efficiency of these meetings: 1. What percentage of discussion time leads to decisions or action items? 2. Which topics consume the most time relative to their importance? 3. Are there topics that could be handled async (via document or Slack) instead of in a meeting? 4. How often does the meeting run over its scheduled time? 5. What is the average number of action items per meeting?"
Best Practices for Transcript Quality
Speaker Attribution Matters
Transcripts with speaker labels are dramatically more useful than those without:
WITH SPEAKER LABELS (good): "Sarah: I think we should delay the launch by two weeks. Tom: I disagree. The marketing campaign is already scheduled. Sarah: Can we decouple the marketing from the product launch?" WITHOUT SPEAKER LABELS (poor): "I think we should delay the launch by two weeks. I disagree. The marketing campaign is already scheduled. Can we decouple the marketing from the product launch?"
Without labels, NotebookLM cannot attribute action items to specific people or track who made which argument.
Handling Poor Audio Quality
If your transcript has errors due to poor audio:
- Review and correct critical sections (decisions, action items) before uploading
- Add a note at the top: “Audio quality was poor. Names and technical terms may be transcribed incorrectly.”
- For critical meetings, ask a participant to review the transcript for accuracy
Supplementary Notes
If important context was shared via screen share or whiteboard (which transcripts miss), add a manual note:
"[Note: During this meeting, Sarah shared a diagram showing the proposed architecture. The three-tier approach was the one the team agreed to pursue. Tom's concerns were about the data layer, not the API layer.]"
These manual additions fill gaps that audio transcription cannot capture.
Frequently Asked Questions
Can I use any transcript format?
NotebookLM accepts Google Docs, PDFs, text files, and copied text. Most transcription tools export to at least one of these formats. For best results, use plain text with speaker labels and timestamps.
How many meeting transcripts can one notebook hold?
NotebookLM supports up to 50 sources per notebook. For a weekly meeting, that is roughly one year of transcripts. If you need more, create a new notebook for the next period and keep the old one as an archive.
Is my meeting content private in NotebookLM?
NotebookLM is a Google product. Your content is stored in your Google account and is subject to Google’s privacy policy. For meetings containing highly sensitive information (M&A discussions, legal strategy, personnel decisions), review your organization’s data handling policies before uploading.
Can NotebookLM generate audio summaries of meetings?
Yes. NotebookLM’s Audio Overview feature can generate a podcast-style summary of your meeting transcripts. This is useful for stakeholders who prefer listening to reading.
How does this compare to dedicated meeting AI tools (Otter, Fireflies)?
Dedicated tools are better at real-time transcription and live meeting integration. NotebookLM is better at post-meeting analysis, cross-meeting intelligence, and deep querying. The ideal workflow uses a dedicated tool for transcription and NotebookLM for analysis.
Can multiple team members access the same notebook?
Yes. NotebookLM supports sharing notebooks with other Google accounts. Share the meeting notebook with your team so everyone can query it independently.
What if the transcript is very long (2+ hour meeting)?
NotebookLM handles long documents well. Upload the full transcript — its large context window can process hours of meeting content. For very long meetings, consider splitting into segments by agenda topic for easier querying.