Welcome to Prompt, Tinker, Innovate—my AI playground. Each edition gives you a hands-on experiment that shows how AI can sharpen your thinking, streamline your process, and power up your creative work.
This week's playground: Turn one source into a full briefing kit with NotebookLM
You just got a 40-page industry report. Your boss wants the key takeaways by end of day. Your team needs a deck for Friday's meeting. And someone on the leadership team will never read it—they want a summary they can listen to between calls.
That's three deliverables from one document. On a normal day, that's a few hours of reading, summarizing, reformatting, and building slides.
Google's NotebookLM does all of it in about five minutes.
Why this matters
Most people know NotebookLM as "that tool that makes the AI podcast." And yes, the Audio Overview feature is impressive. But it's one of about a dozen outputs NotebookLM can generate from your source material—and it's not even the most useful one.
Here's what most people don't realize: NotebookLM is the strongest tool across all major AI platforms when you need answers grounded in specific documents. It doesn't search the web. It doesn't hallucinate from training data. It only works with what you give it. That constraint is the feature.
I wrote about this in The AI Feature Map two weeks ago—NotebookLM wins on source document volume and output variety. Today, I'm going to prove it.
Use case spotlight: The briefing kit
Picture the workflow most of us actually run when a new report—an analyst report, a competitive teardown, a 50-page policy doc—drops:
- Skim the report. Maybe highlight some things.
- Write a summary for leadership.
- Build slides for the team meeting.
- Send the report to people who won't read it, knowing they won't read it.
NotebookLM collapses all four steps into a single session. You upload the report, set a custom instruction to focus the AI on what matters to your team, and then generate:
- A mind map to see the full landscape of topics before you read a single page
- A briefing doc that pulls out the key findings and recommendations
- A slide deck ready for your Friday meeting
- An audio overview for the person who processes information better by listening
One source. Four outputs. Each shaped by your custom instruction so they all tell the same story.
Your AI experiment: Try this prompt
👉 Time to tinker: Grab something you've been meaning to read. A vendor proposal, a quarterly business review, a Forrester report, a board deck, a 60-page compliance doc—whatever's been sitting in your inbox marked "FYI."
Step 1: Set up your notebook
- Go to notebooklm.google.com
- Create a new notebook
- Upload your report (PDF works great; if it's a Google Doc, even better—it stays live-linked)
Step 2: Set your custom instruction
Click "Configure Chat" (the slider icon at the top of the chat box), click "Custom", and paste this:
📝 Prompt (this week, it's a custom instruction):
You are a strategic analyst supporting a marketing leadership team at a mid-size tech company. When analyzing this document, prioritize: (1) findings that directly impact marketing strategy or budget decisions, (2) data points that challenge conventional thinking, and (3) actionable recommendations the team can implement within 90 days. Skip background context the team already knows. Be direct and specific.
Swap in your own role, team, and priorities. The more specific your instruction, the better every output will be.
Step 3: Generate your briefing kit
Now use the Studio panel on the right side to generate each output:
- Mind map first. This shows you everything in the report at a glance. Click any branch to open a chat about that specific topic. Use this to decide what matters before you invest time reading.
- Report second. Click Reports in the Studio panel, then select Briefing Doc (or use the suggested formats—NotebookLM tailors these to your sources).
- Slide deck third. NotebookLM generates a presentation from your sources. New this year—you can now revise individual slides and export as PPTX.
- Audio overview last. Choose your format (deep dive, brief, or debate), set the focus, and let it generate. Share with your team as a "read this report" alternative.
💡 Pro tips
- Upload your brand guidelines as a source. NotebookLM will match colors and fonts in your infographics and slide decks. I tested this—it actually works.
- Use Google Docs instead of PDFs when possible. Google Docs are treated as living documents in NotebookLM. Update the source, and your notebook sees the changes. PDFs are static uploads.
- Delete your chat history between major tasks. The AI can be influenced by previous conversations. If you're shifting from "give me a summary" to "build me a competitive analysis," clear the slate first.
- Try the interactive Audio Overview. Click "Join" during playback and you can interrupt the AI hosts mid-sentence—ask for clarification, redirect the discussion, or quiz them on specific details. It turns a passive podcast into an active study session.
- Stack multiple sources. The real power shows up when you upload 5–10 related documents—competitive reports, internal strategy docs, market data. NotebookLM synthesizes across all of them, and every answer cites which source it pulled from.
What did you discover?
NotebookLM sits in an interesting spot in the AI landscape. It's not the best general chatbot. It's not the best coding tool. It's not the best research agent.
But for one specific job—"I have these documents and I need to extract value from them without hallucinations"—nothing else comes close. And the output variety is unmatched: mind maps, briefing docs, slide decks, audio, video, infographics, flashcards, data tables, quizzes. All from the same source set.
I'm curious: have you used NotebookLM for work? What outputs surprised you? And if you haven't tried it yet—what's the first report you'd upload?
Until next time—keep tinkering, keep prompting, keep innovating.
📩 Not subscribed yet? Hit the button at the top. More experiments coming soon.
⚡️ Start building your own marketing tools ⚡️
The next cohort of Vibe Coding for Marketers (June) is officially open!
Most marketers have great ideas for tools or calculators but get blocked by dev timelines or budget constraints. Vibe coding lets you use natural language to command AI to build real software—no coding experience required.
In this hands-on cohort, we're building:
- Workflow automations: Custom Chrome extensions and app scripts
- Custom lead-gen tools: Think ROI calculators and interactive quizzes
- Microsites: Branded landing pages
Stop being the person who asks for permission and start being the person who builds the solution.
Seats are limited and the last cohort sold out fast. Join us here:
