How I Turned Google’s NotebookLM into an Operational Assistant for My Company

For quite a while, I’ve been thinking about how to improve my company’s revenue — whether by training my team more effectively or by finding smarter ideas to optimize the business.

That’s a common goal for most entrepreneurs… but actually turning it into reality is surprisingly difficult.

The challenge is consolidating information.

Most companies have critical data scattered across different places: long internal documents, disconnected spreadsheets, dashboards that don’t talk to each other, notes in random folders, and knowledge stored only in people’s heads. Even when everything exists, organizing it in a systematic way takes time, and depending on it in day-to-day operations becomes almost impossible.

And for what I wanted — a truly useful AI partner — that fragmentation was the biggest obstacle.

My First Attempts with AI Tools

My first experiments were with ChatGPT, on the most affordable subscription.

It was helpful, but at least at the time, the limitations were clear:

• Memory was weak
• Context had to be re-explained constantly
• Maintaining continuity across sessions was hard

Then came the arrival of MCPs, and I decided to try Claude.

Claude was excellent at what it did, but the downside was also obvious:

• The pricing wasn’t very friendly
• There was a noticeable delay when making MCP queries
• And once again… the familiar limitation: memory

The truth is, context is everything in operations.
Every day you remember something new, discover a new metric, adjust a process, learn a lesson… and feeding all of this into an AI repeatedly is exhausting.

Eventually, I paused the project.

Having AI as an operational ally felt much harder than I expected.

What I Actually Wanted (And What I Didn’t Want)

At that point, I realized something important:

I didn’t want a complex solution.
I didn’t want a local RAG setup, an over-engineered workflow, or some Frankenstein stack that only works if you babysit it.

What I wanted was simple:

✅ A tool where I could say:
“This is my company. This is what we do. These are my numbers. These are my manuals. Now think with me.”

No heavy workflows.
No complicated pipelines.
Just an AI that could understand my business and behave like a strategic assistant.

Then NotebookLM Launched (And I Almost Ignored It)

When Google announced NotebookLM, I saw the news… and didn’t even connect it with my old need.

Honestly, I didn’t care at first.
It felt like “just another AI product.”

But a few days later, I decided to open it.

And immediately, I was impressed.

NotebookLM had incredible capabilities:

• Audio generation
• Summaries and structured insights
• Infographics
• Fast reasoning over source documents

But what really caught my attention wasn’t any of that.

It was one specific feature:

Sources.

NotebookLM wasn’t just “answering like an AI.”
It was answering based on the material I fed it — and it could reference those sources.

That was the missing piece.

My Plan: Build a Business “Consiglieri”

At that moment, I knew what I wanted to do:

I would train NotebookLM to act as an operational advisor — almost like a consiglieri for my company.

But to do that, I needed cold, structured data:

• Google Analytics
• Google Search Console
• MySQL company data (sales, employee performance, operational metrics)
• Manuals, playbooks, business books, training material

I already had an internal table syncing to Google Sheets from a previous project.
What I needed next was to connect Analytics and Search Console into the same flow.

How I Consolidated Everything Into NotebookLM

Here’s what I did:

1. ✅ Synced Google Analytics and Search Console data into BigQuery
2. ✅ Used a native BigQuery connector to bring those datasets into Google Sheets
3. ✅ Organized everything in the same Sheets environment (alongside my MySQL operational data)
4. ✅ Added that Google Sheet as a primary source inside NotebookLM
5. ✅ Attached supporting documents: manuals, books, internal explanations and reference material

Once everything was in place…

NotebookLM became something completely different.

It wasn’t just “an AI chatbot.”
It became a business assistant powered by my real company context.

The Result: It Worked

Suddenly, I could:

• Query sales performance per employee
• Ask for insights based on real traffic and lead behavior
• Generate strategic suggestions grounded in my metrics
• Create infographics for possible business plans
• Use it as a training assistant for my team
• Brainstorm operational improvements with actual data, not generic advice

And most importantly…

The idea worked.

I finally had what I wanted from day one:
A simple AI setup that didn’t require complex engineering — and that could truly understand my business.

Final Thought

NotebookLM might look like a product meant for students, research, and note-taking.

But in practice, if you feed it the right sources, it can become something much bigger:

✅ An operational assistant
✅ A strategic partner
✅ A business advisor that actually knows your numbers

And once you experience that, it becomes hard to go back.