I’m about to lose money showing you this.
I charged a client 10 grand for something you can build for free right now. This AI assistant reads every document in your business. SOPs, contracts, reports. And it answers any question with exact citations.
I built it for a client in under an hour. By the end of this post, you’ll have the same thing running. No coding, no extra fees, no developers.
The Two Tools You Need
I’m Charles Dove, and I run CC Strategic with my partner Chris. I build AI automation systems and share everything on @charlieautomates on YouTube.
Here’s the thing. The secret is stacking two tools together: NotebookLM and Claude Code.
NotebookLM is a free tool from Google. It lets you create notebooks and plug in information. Websites, files, stuff from your Drive, plain text. You can chat about it, ask questions, and create different types of assets. Flashcards, reports, mind maps, slide decks, infographics. My two favorites are slide decks and infographics.
Claude Code is Anthropic’s AI coding assistant that lives inside VS Code. If you haven’t set it up yet, check out my playlist on the Charlie Automates YouTube channel. I have a bunch of setup videos there.
The magic happens when you connect them.
Why This Combo Beats the MCP Server
Now, there is an MCP server for NotebookLM. If you’re familiar with MCP servers, they’re basically a toolbox. You can access another software and use all their tools.
But the NotebookLM MCP server only lets you push text and content to NotebookLM. It can’t pull information back. That’s a one-way street.
The script I’m showing you (called NotebookLM PI) fixes that. You can push information to NotebookLM and pull it back out. You can have it create those assets, download them locally, and use them however you want. All in one place.
That’s the $10,000 product nobody knows exists.
How to Install NotebookLM PI
This isn’t super complicated. Here’s the setup.
- Go to the NotebookLM PI repo on GitHub
- Copy the install command for your operating system (Windows or Mac)
- Paste it into your integrated terminal inside VS Code
- Hit enter and let it run
- A browser window opens for you to sign into Google
- Sign in, come back to VS Code, and hit enter again
- It authenticates and saves everything
That’s it. You’re connected.
Once it’s installed, you just chat with Claude. Say “Hey, can you take a look at my NotebookLM account?” and it’ll show you all your notebooks.
Use Case 1: Sales Reconnaissance and Presentations
This is the one that made my client pay 10 grand. And it’s so simple it hurts.
I had a sales call with a prospect. I dropped the transcription from that call into a notebook through Claude Code. I said, “Hey, create a notebook for this client.” Done.
Now here’s where it gets good.
I asked Claude to tell me about the notebook and the conversations it contained. It pulled up full context on the conversation. Different turns, key points, everything with citations.
Then I asked it to create a mind map. It built a custom HTML mind map right there on my computer. Not some basic text outline. A full interactive mind map.
I had it create an entire slide deck for the next presentation based off the sales conversation. Downloaded locally. Ready to use.
Building a Client Profile Over Time
Here’s the real power. Let’s say you have a second call with that prospect. You feed the transcript of the second call to Claude. Ask it to update the notebook. Now you have a running profile for that prospect.
Call number two showed up right in the notebook. The AI gave me context on what changed between the two calls. How expectations shifted. How we might pivot our proposal.
You could even create a third presentation for the close call based on everything from calls one and two. The AI already has all the context.
The key is how you organize things. Name your notebooks and sources clearly. That’s how Claude knows which notebook to use for what. I’d suggest creating a separate notebook for each individual prospect. It keeps things clean.
Token Savings
Here’s the part nobody talks about. NotebookLM costs you no extra tokens. Zero.
You have an extra AI assistant doing research and creating assets, and it doesn’t eat into your Claude Code token budget. NotebookLM handles the heavy lifting. Claude Code just orchestrates.
All you’re paying for is your main Claude Code subscription. That’s it.
Use Case 2: Sales SOPs and Team Training
The second way I’m using this personal assistant is for SOPs.
I have a Notion document with my sales processes. Specific processes for manual outreach. Specific softwares and forms my team needs to use. Standard operating procedure stuff.
Here’s what I did. I took all that SOP information, uploaded it to Claude, and asked it to create a new notebook for that part of my business.
Now instead of my employees or new hires or salespeople bothering me or my sales manager with questions, they can reach out directly to Claude through Claude Teams. They get the information they need instantly.
How It Handles Software Links and SOP Videos
This is a detail that matters. When you pull your SOPs in, if you have specific links for software tools, they stay hyperlinked. When your team asks Claude a question through the NotebookLM notebooks, it gives them those links.
So if the answer involves a specific software or an SOP training video, those links are clickable right in the response. Your team doesn’t have to dig through a shared drive.
Building an Objection Sheet From Real Calls
Want to take it further? Feed your actual sales call transcripts into the SOP notebook. Claude can create an objection sheet from real conversations.
So whenever your salespeople run across typical objections, those objections and the best responses can be fed to them via AI. Instead of you being the one who answers every question.
That’s how you scale yourself out of the day-to-day.
Use Case 3: Client Onboarding Notebooks
The third use case is client onboarding. And it goes hand-in-hand with SOPs.
You have an onboarding flow for new clients. Teach those SOPs to a notebook so your team can understand exactly how to onboard each client.
Same process. Ask Claude to create a new notebook. Feed it the SOP documents, video links, software links, everything.
Here’s the twist. You can add individual client information into the notebook and update it as the engagement progresses.
How This Works Day to Day
Let’s say your support person is hopping on a technical call with a client. Your tech team updated the notebook after the last call. Your support agent just wants to know where the project stands.
They chat with the AI. They get an instant update on where the project is at. No Slack threads to scroll through. No waiting for someone to reply.
You can even use this for onboarding internal employees. New hires can ask the AI about processes, tools, and standards instead of interrupting senior team members.
The Real Secret: It’s Just Organization
At its core, this is just an AI assistant. You’re storing data in NotebookLM. Google’s AI sifts through it. Claude Code orchestrates the whole thing.
The important thing is how you organize the information. How you name your sources. How you name your notebooks. That’s what makes Claude recognize which notebook to use for what purpose.
No right or wrong way to do it. It’s really just how you want to organize your business knowledge.
Step-by-Step Setup Summary
Here’s the quick-hit version:
- Install Claude Code in VS Code (check the YouTube playlist if you haven’t)
- Install NotebookLM PI using the terminal command from the repo
- Authenticate with Google when the browser window opens
- Create your first notebook by telling Claude what you want to organize
- Feed it documents like sales transcripts, SOPs, onboarding guides
- Ask Claude to create assets like mind maps, slide decks, or summaries
- Pull context whenever you need it by chatting with Claude
Everything downloads locally. Everything stays organized. And NotebookLM doesn’t cost you a dime beyond your Claude subscription.
Why This Is a $10K Product
Let me be real. I charged a client 10 grand for this setup because it replaced multiple tools and saved their team hours every week.
The prospect research alone used to take 30 to 45 minutes per call. Now it takes 2 minutes.
Building presentations from sales calls? Used to be an hour of work minimum. Now Claude does it while you grab coffee.
Training new salespeople on objection handling? Used to mean shadowing calls for weeks. Now the AI teaches them from real conversation data.
This is a part of my personal setup. This is my secret sauce. And now you have it too. Welcome to the club, fam.
FAQ
Is NotebookLM really free?
Yes. NotebookLM is a free tool from Google. There’s a Plus tier with extra features, but the free version handles everything we covered in this post. Combined with your Claude Code subscription, you’re not paying anything extra.
Do I need to know how to code for this?
No. The entire setup is done through conversational commands with Claude Code. You install one script through the terminal, sign into Google, and then just chat. If you can type a sentence, you can do this.
What’s the difference between the NotebookLM MCP server and NotebookLM PI?
The MCP server only pushes information to NotebookLM. It can’t pull data back or create assets. NotebookLM PI does both. You can push documents in, create mind maps and slide decks, download them locally, and pull context back into Claude Code.
How many notebooks can I create?
There’s no practical limit for most business use cases. I’d recommend one notebook per prospect, one for each SOP category, and one for each client engagement. Keep them focused and clearly named.
Can my team access the notebooks through Claude Teams?
Yes. When you set up Claude Teams, your team members can chat with Claude and access the NotebookLM notebooks. They get the same citations and context you do. That’s how you replace yourself as the bottleneck for answering process questions.
Does this use my Claude Code tokens?
NotebookLM handles the heavy processing, so it saves tokens compared to having Claude Code sift through all your documents directly. The only tokens used are the Claude Code conversation itself. The research and asset creation happen on Google’s side.
What types of assets can NotebookLM create?
Flashcards, reports, mind maps, slide decks, infographics, quizzes, and more. My two favorites are slide decks (for sales presentations) and infographics (for client deliverables). All of them download locally to your computer.
Want more setups like this? Join CC Strategic AI on Skool where I share Claude Code workflows, templates, and the exact systems I use to run my agency.
If you want to go deeper, faster, work with me 1-on-1. We can build your AI assistant setup together in a single session.
And if your business needs a full AI automation system, book a call with CC Strategic. That’s what we do.