How to turn 50 finance PDFs into 1 searchable brain in 20 minutes


If you want to become an AI Finance Leader

I'm launching the best opportunity to transform your career and become the AI Finance Leader of your company.

It's a 6-week hands-on live bootcamp which will give you all the practical knowledge you need to make financial analysis with AI, automate your reporting and forecast with AI.

If you want to invest in your career, click below:

Last Monday, a head of tax told me she found a way to consolidate all the recent tax updates across multiple documents and websites (which would normally take hours).

She then found a way to automatically create a podcast to listen to these updates!

Not only was this a more digestible way for her to consume all of this information, she could also multitask or listen on her commute also.

But, tax is just one example. How many other times do you work with messy documents in different formats? Analysis, reporting, SOPs. It’s a never ending list.

Normally, connecting the dots between documents will take up your weekend.

You read a contract, then check a spreadsheet, then read a tax update, and you try to hold it all in your head.

But your job is not to keep information in your head. It is to uncover insight that other people cannot see (and learn faster than anyone else).

Today, I will teach you how to do this. And it will take you no more than 20 minutes.



Data Blindness

In finance, you forget just how much data you are working with every day.

You are really good at remembering where everything is. But generating reports and conducting analysis from 50 different PDFs and spreadsheets feels impossible.

Imagine you get asked to do some competitor and benchmarking.

You download four annual reports and each one is 80+ page long.

You open the first PDF, read through 8 pages, and highlight a few things (maybe you also create some of your own docs for notes).

Then you open the second PDF. The information you need is on page 31. But, it’s in a different format, and categorised in a different way.

Before you know it, you’re no longer analyzing, you’re just re-formatting the data so it makes sense.

You’ve lost half a day already, and you’ve not even looked at any financial statements yet!

But the bigger problem is you are missing really important things as well.

This is because you can't see patterns across different documents in different formats, all at the same time.

Maybe the annual reports from Company A and Company B both have a key insight you didn’t account for.

But you didn't notice because you read them two days apart and became blind to it.

Your brain cannot hold that much context.

Your eyes cannot see across 10 documents at once.


AI Hawkeye

So here's the better way.

Google NotebookLM.

NotebookLM is not the same as ChatGPT, Gemini or Copilot.

It acts more like a secure research assistant that only knows what you teach it.

You're not asking it to make things up (generate) or hallucinate. You're giving it SPECIFIC documents and asking it to connect the dots for you.

It will give you an advantage because:

  1. It includes precise references. It cites specific rows, paragraphs, and page numbers. Every answer includes the source reference so you can verify it easily.
  2. It can handle multiple large documents easily. Upload 5 contracts, 3 spreadsheets, and 2 x 100 page PDFs. Ask it to find conflicts between them and it sees all of them at once.
  3. It's built for analysis, not generation. This isn't about writing reports from scratch. It's about synthesis. How do these documents relate to each other? What are we missing? Where do they contradict? What can I learn from them?

If your target is 1 data point in a jungle of thousands of words and numbers that are super difficult to see, NotebookLM is your best chance of hitting it.

So, how do you use this?



How to use NotebookLM in 3-Steps

Step 1: The Upload (Building the Brain)

Create a new Notebook for a specific project. Best practise is to have one notebook per project.

Examples:

  • For Strategy: Upload your competitor's last 3 investor presentations (PDF) + your internal strategy memo.
  • For Operations: Upload your monthly variance report (Google Sheet) + the raw export of sales data.
  • For Compliance: Upload the new regulation (PDF) + your current policy document.

You can also bring in external data (websites, YouTube) to enrich your research even more.

Step 2: The "Synthesis" Prompt

Don't ask for a summary. A summary just tells you what you already have. Ask for a Synthesis. How the documents relate to each other.

Try these prompts for the 4 key areas:

1. Strategic:

"Compare the 'Risks' section of Competitor A's annual report (Source 1) with our own Risk Register (Source 2). Where are they seeing threats that we are missing?"

2. Operational:

"Look at the Q3 Expense Report (Source 1). Identify the top 3 categories where we exceeded budget, and cross-reference them with the Departmental Operating Plan (Source 2) to see if these were planned initiatives or accidental overspends."

3. Regulatory:

"Review the new Tax Compliance Guide (Source 1). Highlight exactly which clauses contradict our current Internal Finance Policy (Source 2)."

4. Tech:

"Based on the ERP User Guide (Source 1), draft a step-by-step 'How-To' checklist for our team to process a purchase order, simplifying the technical language."

The important thing is you can always see where the information was taken from. That way, you can trace and check different resources at the same time.

Step 3: The Output

You will be shocked by how much NotebookLM can do.

It doesn't just read the text, you can use it to find the connections and themes between multiple documents in different formats.

Plus you can also get it to:

  • Produce an audio podcast
  • Generate a video
  • Create a mind map
  • And much much more

NotebookLM will hallucinate less when compared to LLMs like ChatGPT, as it's working from specific data.

Just remember, it's only as good as the data.

So always check the data, and the output.

Fore example, if you generate a slide deck visual, numbers and charts will need to be double checked.

The citations will be accurate. But the bar chart might look a little weird, or the font might be off. That is fine.

Your job is to find the insight. You can fix the font in PowerPoint (or Slides) later.

The value here is that you just turned 12 hours of reading and cross-referencing into a 15-minute review process.


5 More Use Cases to Inspire You

Here are some other ways you can use NotebookLM.

1. Board Report Preparation Upload last quarter's management pack + this quarter's actuals + your CEO's strategy memo. Generate a slide deck highlighting the variances.

2. Policy Change Communication Upload your new expense policy + last quarter's violations. Create a video overview explaining what changed. Share it with your team instead of sending another email (that nobody reads).

3. Financial Statement Analysis Upload 3 years of competitor annual reports. Create a mind map showing how their revenue streams and margin structure have evolved.

4. Compliance Cross-Check Upload the new ESG standard + your current disclosures. Generate flashcards showing what you're missing for your next compliance meeting.

5. M&A Due Diligence Upload the target's financials + contracts + operational reports. Create an infographic summarizing key risks and opportunities.


The one thing to remember

With NotebookLM you can have both speed and accuracy across all your finance documents (even if they're a complete mess).

So, take 10 minutes this week, and pick the biggest documents you dread reading.

The ones you've been avoiding because you know it's going to be a nightmare to cross-reference them with everything else.

Upload them to NotebookLM with one or two related files. Then try one synthesis prompt.

See what it finds. See if it connects something you missed.

It’ll quickly turn into your unfair advantage in uncovering insight that other people cannot see.

Best,

Your AI Finance Expert,

Nicolas


P.S. - What did you think of this edition? Hit reply and let me know (I read all replies).

P.P.S. - If you want to hear my discussion where Josh Aharonoff (Your CFO Guy) and I talk about NotebookLM, you can watch the YouTube video here.


Behind The Scenes

Recently, I helped the finance team of AUTODOC to upskill 50 people on AI in 30 days.

I love running these corporate training sessions. From beginner tips to building use cases, to helping the business save hours of work, and improving the value provided by finance.

Reply to this email if you are interested to run a similar workshop with your team (remote and in-person) in 2026

.

Tips & Insights on Finance & AI

Join 270,000+ Professionals and receive the best insights about Finance & AI. More than 1 million people follow me on social media. Join us today and get 5 goodies from me!

Read more from Tips & Insights on Finance & AI
90% faster presentations in 3 steps

If your goal is to be an AI Leader in 2026: I'm launching the best opportunity to transform your career and become the AI Finance Leader of your company. It's a 6-week hands-on live bootcamp which will bring you from beginner to expert in AI for Finance and will allow you to automate, analyze and forecast with AI. If you want to seize the AI opportunity in 2026, this is for you: Join the Waitlist It’s 4PM the day before your board meeting. But, you’ve been so busy this month that you’ve not...

AI Finance Accelerator - Waitlist now open If you want to get ahead with AI whilst other finance pros are still 'experimenting' - my AI Finance Accelerator is for you. After training 10,000+ finance professionals, you won’t get ‘just another course’. You will actually implement AI in your finance team in 6-weeks. Early access pricing: First 50 only (I expect these to be gone in the next days). So you do not miss out - Apply by completing the form behind the link below. Join the Waitlist for...

January 2026 Waitlist Now Open If you want to get ahead with AI whilst other finance pros are still 'experimenting' - my AI Finance Accelerator is for you. After training 10,000+ finance professionals, you won’t get ‘just another course’. You will actually build: AI-enabled reporting (not dashboards you'll never use) Finance agents (not chatbot demos) Python/ML forecasting (not Excel macros) Real automation you can run as soon as the cohort ends Investment: 6-8 hours/week: If you're too busy...