Hi there!
Welcome to the 26th edition of Work in Beta.
In this edition, we get into NotebookLM: what it’s actually for, when to reach for it instead of ChatGPT or Gemini, and how to set one up that earns its place in your week.
One of our free AI Starter Kits is built entirely around Gemini, and it covers NotebookLM too. If today’s edition makes you want to try it for yourself, that’s the fastest way in: open the Gemini kit and you’ll have something working in 10 minutes. Start here.
So, let’s dive in!
IF YOU ONLY HAVE 2 MINUTES
Image Credits: ChatGPT / Work in Beta
THE ‘HOW TO’ PLAYBOOK
When to Use NotebookLM Instead of ChatGPT.
The answer you need is already in your files. It is in that long contract. Or last month’s report. Or your meeting notes. You know it is in there. You just can’t find it fast enough.
So you copy a few pages into ChatGPT and ask. It gives you an answer. But part of that answer came from your file, and part of it the AI made up. Both parts sound the same. Now you have to work out which part is real.
Why does this happen? ChatGPT, Claude, and Gemini have read almost the whole internet. So when your file is missing something, they fill the gap on their own. A lot of the time, that helps. But sometimes you want the opposite. You want an AI that looks only at your files and nothing else.
That AI is NotebookLM. Think of it as a Source Room. You choose what goes in the room: your reports, your contracts, your notes. Then it answers only from what is in the room. It even shows you the exact line it used. Nothing from the internet. Nothing made up. Just your stuff, ready to answer you.
The One Question to Ask
A normal chatbot answers from everything it has read. A Source Room answers only from what you put in it. That is the whole point.
It also tells you when to use it. Forget the names of the tools. Just ask one thing:
Do you want the answer to come from your own files? Or from everything the AI knows?
If it should come from your files, use NotebookLM. Like: “What did the client agree to in these three contracts?” The answer is sitting in your files. You just want it found and quoted.
If it needs the AI to think, plan, or search the web, use Gemini, Claude, or ChatGPT. Like: “Write me a refund policy.” That answer is not in your files yet.
Quick check: if a made-up answer would annoy you, you want the Source Room.
We have talked about this before, in Are You Using AI on the Wrong Work? and Claude Chat or Cowork? How to Tell Which One Your Work Needs. The trick is always the same. Name the work first. Then the right tool is easy to see.
Step 1: Put the Right Things in the Room
A Source Room is only as good as what you put in it.
Use one notebook for one project. Not one giant notebook for everything. Each notebook only knows its own files. That sounds like a problem. It is actually the best part. It keeps the answers clean.
Add only the files that matter. You can add up to 50 files in one notebook on the free plan. But more is not better. Ask about the 3 files that matter, and the answer is sharp. Ask across all 50, and the answer gets fuzzy. Most of the time, a few good files is all you need.
Check what really went in. This part trips people up. What you see in a file is not always what NotebookLM saved:
A web link comes in as text only. No pictures. No videos. Nothing behind a paywall.
A YouTube video comes in as the words only.
A Google Doc comes in without its comments or footnotes.
So before you trust a notebook, look at what it actually pulled in.
Step 2: Put the Room to Work
Most people only ask questions. But NotebookLM can also make things from your files. A short audio show. A one-page picture. Even slides.
The smart move is to build a room for the work you do again and again. Here are four:
Prepare: a room for a meeting. Add your notes, the plan, and the client’s last few emails. Ask: “What is still open? What does not match up? What should I be ready for?” Now you have a quick brief instead of reading six files at 8am.
Propose: a room for winning work. Add the request, two proposals that won before, and your best examples. Ask it to match each ask to proof you already have, and to point out what is missing. Do not ask it to write the whole thing. Ask it where your proof is, and where it is not.
Train: a room for new people. Add your how-to guides and a few recorded walk-throughs. New folks ask simple questions and get answers that point right back to the guide. They stop asking you.
Catch Up: a room for reading. Add reports and articles. Turn them into a short summary, or a little audio show for the drive home.
When you make an audio show, pick the style that fits: Deep Dive for a full brief, Brief for a fast catch-up, Critique to poke holes in a draft, Debate to hear both sides before you decide.
Step 3: Know When to Leave the Room
The Source Room is for working with your files. When the job grows past your files, move with it.
Take it to Gemini. Your notebooks now link up with Gemini. Same files. Same setup. The difference is simple. NotebookLM is the quiet room with just your files. Gemini is the workshop: it has your files too, plus web search and more tools. So you read and ask in NotebookLM. Then you draft and plan in Gemini. (Two honest notes: the audio shows and pictures only get made inside NotebookLM. And the two apps treat your data by different privacy rules.)
Also know when to stop:
Hard math. It can read a spreadsheet and still count it wrong. Do real sums in Sheets or Excel.
Big decisions. The plan and the final call are yours. It cannot find an answer that is not in the room.
Sources you did not open. A source tells you where it looked, not that it got it right. It can point to a real page and still read it wrong. Some tests caught it making up quotes and mixing up dates. So open the source and read the line. Boring, but it works.
The Mistakes We See People Make
Using it like ChatGPT. You ask for an opinion and get a flat answer. So you think the tool is weak. It is not. You asked a librarian to be a consultant.
Dumping in everything. Fifty files feel safe. They make answers fuzzy. Add only the files that matter.
Trusting a source you never opened. The link is not proof. Open it and read the line.
Using it for hard math. It sounds right and counts wrong. Keep the math in a spreadsheet.
Forgetting what did not load. Comments, footnotes, paywalled pages, videos. If it did not go in, it cannot come out.
Final Thought
The answer to most of your file work is already in your files. The only question is whether you can get to it. Get to it without reading everything again. And without trusting an AI that fills the gaps when your files fall short.
A Source Room does both. So build one this week. Pick a real project. A client. A proposal. A task you keep explaining. Put its files in one notebook. Ask the question you would normally dig for. Then open one source and check it. That is the whole skill.
WORK WITH US
Build With Us
Most professionals know AI can do more for them. The gap isn’t awareness - it’s knowing where to start, what to change, and how to make it stick.
That’s what we work on through Work in Beta.
For individuals, we run working sessions, not teaching sessions. You bring a real problem from your actual work; we build the solution with you, live. You’re borrowing our learning curve instead of grinding through your own. You walk out with something that works and the muscle to keep going. When you get stuck later, we’re a message away.
You probably have a version of at least one of these:
A task you redo from scratch every time, even though the steps never change.
Something you’re good at that’s only ever lived in your head, never as a tool you can actually use.
A workflow you started automating and gave up on halfway.
Bring that. We’ll build it with you.
For organizations, AI adoption is a people problem, not a technology problem. Your teams have the tools, what’s missing is the translation layer between AI capability and daily work. Which processes to redesign, which habits to break, how to build genuine fluency - not just awareness. We help close that gap through hands-on training, process redesign, and deep adoption engagements. Not advisory, forward-deployed.
If any of this resonates, email us at [email protected] / [email protected] and we will figure out how to work together.