NotebookLM just replaced a data team

I fed it 300 messy files

June 21, 2026

In partnership with

I have spent more hours than I want to admit staring at a spreadsheet, scrolling for the single broken cell that throws off the whole sum. So when a quiet update shipped for Google's NotebookLM last week, I almost scrolled past it. Then I watched an AI pro walk through what it can now do, and I had to put my coffee down.

She used it to replace an entire data analysis team. Not "summarize a PDF" replace. Actual "find the leak in your business and tell you why it is leaking" replace. I watched the breakdown twice because the third step is where it stops being a search box and starts being a thinking partner.

Here is what changed, and exactly how I am going to run it on my own numbers tonight.

LLM traffic converts 3× better than Google search

58% of buyers now start their research in ChatGPT or Gemini, not Google. Most startups aren't showing up there yet.

The ones that are get cited by the AI tools their buyers, investors, and future hires already use. And they convert at 3×.

Download the free AEO Playbook for Startups from HubSpot and get the exact steps to start showing up. Five minutes to read.

*Ad

What actually changed under the hood

NotebookLM is now running on the latest Gemini, but the engine swap is not the headline. The headline is the new upload limit. You can feed it up to 300 of your own sources at once.

And it does not care how messy they are. Raw CSV exports. Unstructured text dumps. A profit and loss statement. A pile of support tickets. Once your files are in, the AI stops being a chatbot and starts acting like a data analyst that never sleeps. It digests the rows and columns and turns them into clean charts, visual graphs, infographics, even polished slides you could drop into a deck.

The piece that matters: you do not write a single formula. You ask plain-English questions and it does the digging.

The cupcake business that proved it

The demo ran on a fictional cupcake company, and the detail is what sold me. She uploaded seven spreadsheets tracking everything from weekly Instagram posts and email send volumes to individual Stripe transactions, one-time fees sitting next to monthly subscriptions. She even dropped in a full profit and loss statement with cost of goods, shipping, software, and payroll.

Old way: weeks of pivot tables and formulas to make sense of that. New way: she just asked questions and let the tool find the signal in the noise.

"Which product has the highest refund rate?"

It read the Stripe data and pointed straight at a baking course that accounted for nearly all the refunds. Thirty seconds. No filtering, no formulas.

PRDs by voice. Bug reports by voice. Ship faster.

Dictate acceptance criteria and reproductions inside Cursor or Warp. Wispr Flow auto-tags file names, preserves syntax, and gives you paste-ready text in seconds. 4x faster than typing.

*Ad

The twist that actually blew my mind

Here is step three, the part that reframes the whole tool. The leverage is not the math. The leverage is that it cross-references your clean financial data against your messiest unstructured text.

She selected the Stripe refund data and her raw customer support inbox together, then asked a single question: why are these people asking for their money back? Instead of her reading hundreds of angry emails, the tool read all of them instantly and grouped the complaints by theme.

The verdict: customers were buying the course and never getting the automated access link. It also flagged broken password resets and videos that would not load. A broken technical workflow that might take a human weeks to spot, surfaced in minutes.

Then it went one further. Reading through the general product complaints, it spotted a recurring, unprompted demand for gluten-free and dairy-free options. That is not a bug report. That is a validated new product line, handed over for free, pulled out of emails nobody had time to read.

How to run this on your own business

I am not retelling this to be impressed. I am retelling it because you can copy it in about twenty minutes. Here is the play:

  1. Gather your scattered data. Download your transaction history from Stripe or your payment processor, export your latest ad results, and grab a big dump of your support emails. Messy is fine. Messy is the point.

  2. Make a new notebook and dump it all in. Drop every spreadsheet at once. Give it a moment to digest. What you just built is a centralized brain for the business.

  3. Select only the sources for the question you are asking. Want to know if Instagram actually drives sales? Check the box on your marketing tracker and your sales data, and nothing else. The AI ignores the rest and hunts only for that one correlation. This is the trick most people miss.

  4. Write the question like you would say it out loud. "Build me a one-page monthly profit and loss summary." "Rank my ad campaigns best to worst by return on ad spend." No syntax, no functions.

  5. Generate the final asset. Use the built-in studio to turn the answer into a slide deck, a downloadable PDF, or an audio summary you can send your team while you make coffee.

Global HR shouldn't require five tools per country

Your company going global shouldn’t mean endless headaches. Deel’s free guide shows you how to unify payroll, onboarding, and compliance across every country you operate in. No more juggling separate systems for the US, Europe, and APAC. No more Slack messages filling gaps. Just one consolidated approach that scales.

*Ad

Three prompts I am stealing immediately

Audit the ad spend. Before touching competitors, she pointed it at her Meta ads and asked for total spend, cost per lead, and return on ad spend per campaign. It ranked them instantly and exposed a brand-awareness campaign burning thousands for almost nothing, while a free-recipe-PDF campaign quietly printed money. That is a budget reallocation you can make this week, no data scientist required.

Find the gap your competitors left open. She scraped hundreds of public reviews from rivals on Trustpilot and Google, dropped the text file in, and asked for the single biggest strategic gap she could exploit. The answer: competitors had great flavors but terrible shipping and support. Instant operational target to beat.

Diagnose your churn. She asked when people cancel, and the tool found most users drop off right after month one. Cross-referenced against support tickets, it revealed the real reason: too expensive to keep monthly, and the portion sizes felt too large. It even proposed the fix, a lower second-month price and a smaller box option. The model did not just spot the problem. It wrote the retention offer.

Where I would start tonight

The thing I keep circling back to is not the charts. It is the cross-reference. For years the numbers lived in one tool and the customer's actual words lived in another, and connecting them was a job nobody had time for. That gap is where the money quietly leaks out. NotebookLM just closed it.

So here is my tonight move, and yours if you want it. Open NotebookLM, make one notebook, and upload the single messiest export you own, the one you have been avoiding. Pair it with your support emails. Ask it one honest question: where am I losing money and why. Then sit with what comes back. I think you will find at least one thing you did not want to know, which is exactly the thing worth fixing first.