Forecasting is difficult, but making good decisions from your finance data is an even bigger challenge.

Especially if you cannot afford an expensive FP&A tool (and even when you have one sometimes).

Your business has 20 variables. And they all matter. Marketing spend, sales volume, customer count, pricing, operating costs, seasonality. The list goes on.

And here's the thing you miss. When marketing spend goes up, sales volume changes. When sales volume changes, operating costs shift. When costs shift, your margins move. Everything connects.

But you analyzed them one at a time. Revenue in January. Costs in February. Marketing in March.

So you had 20 separate spreadsheets with reasonable conclusions. But you never saw how they connected.

And this is the problem. Your brain can't process more than 2 or 3 variables at once, and, as a result, this creates big gaps in your decision making. We're just not built for handling lots of data.

Machine learning is.

​

​
Sequential Analysis

This is the problem with 'Sequential Analysis'.

You analyze revenue in one spreadsheet. Then costs in another. Then marketing in a third. Each analysis happens separately, in its own timeline, with its own conclusions.

They never talk to each other.

So when your CEO asks: "Should we increase marketing spend in Q3?" you're stuck.

You know marketing drove revenue growth last quarter. But you don't know if it's profitable growth. Because you analyzed marketing separately from the cost of fulfilling those sales. Separately from support team capacity. Separately from cash collection timing.

You need two weeks to "pull the data together."

Meanwhile, your competitor already doubled their marketing budget and grabbed market share.

Or worse – you DO answer the question. You say yes based on incomplete information. Marketing scales up. Revenue grows. And six months later you're explaining to the board why margins collapsed and cash got tight.

Not because you made a bad decision. Because you made it with incomplete information.

This is what happens when you analyze variables one at a time instead of seeing how they interact.

​

​
Pattern-Based Analysis

A much better way is 'Pattern-Based Analysis'.

Instead of analyzing one variable at a time, you analyze all of them together. Marketing spend AND sales volume AND operating costs AND seasonality. All at once.

Machine learning looks at how they interact. It spots patterns you'd never see manually because your brain simply cannot process that many variables simultaneously.

Let me show you a real example.

Kraft Heinz:

Their manufacturing team had 100+ factory variables to track. Newest machines, fastest machines, most people, least people, most products, least products.

Sequential analysis would mean checking each one individually. Which takes weeks.

So they used machine learning to analyze all 100+ variables simultaneously. They identified the 3 most important drivers that actually determine operational performance and remove the noise.

This insight came from Christian Martinez (an expert in the AI Finance Club), who showed how machine learning can cut through complexity to find what is most important, whilst avoiding the noise.

And you can do the same thing in your finance team. Here are some ways:

So, let me show you how you can start using this right now.

​

How to Run Machine Learning in 6-Steps (No Data Scientist Required)

Here are the tools that you will need:

• An LLM chatbot – ChatGPT, Microsoft Copilot, Google Gemini, or Claude all work
• ​Google Colab - This is a Jupyter notebook to run Python code (free, no installation)
• Your data exported as CSV

Important: If you're using professional company data, you'll need a professional license (ChatGPT Business/Enterprise, Copilot for Microsoft 365, etc.) to keep your data secure. The free versions work for learning with dummy data.

Here's the workflow:

Step 1: Start Simple – Export One Dataset

Don't try to analyze everything at once. Pick one or two things you want to understand better.

Examples:

• Monthly revenue for the past 2 years
• Customer churn data with 5-10 key attributes
• Operating expenses by category for 12 months

Export them as CSVs. Keep it manageable – 2-3 years of monthly data or a few thousand rows maximum.

Why? If you upload 50 variables and 10 years of data into ChatGPT right now, you won't understand what comes out. Start small, learn the process, then scale up.

Step 2: Upload to Your LLM and Describe Your Goal

Upload your CSV to ChatGPT (or whichever tool you're using), with thinking mode turned on.

Then describe what you want to do.

Here’s the prompt I used for the above video.

ChatGPT (or other AI) will think through your request and look at the data.

It should also explain why that model fits your situation.

This is useful so that you can learn as you go.

Step 3: Run a Preliminary Analysis in ChatGPT

Before jumping to generating the code, ask ChatGPT to run a quick preliminary analysis right there in the chat.

Again, only do this if you have a professional / secure license for your AI tool. Otherwise, use dummy / anonymised data before running it with your actual data in Google Colab.

This takes 2 minutes and helps you understand if you're on the right track before you start writing code.

If the preliminary results don't make sense, adjust your approach now. Don't wait until you're in Colab.

Step 4: Generate the Python Code

Once you're happy with the preliminary analysis, ask ChatGPT:

ChatGPT will give you the code. Copy it.

Step 5: Open Google Colab and Run the Code

Now we're going to productize what you just tested (so you can do this regularly without having to use AI every time).

Go to colab.google.com.

You don't need to know how to code here. You just paste the code ChatGPT gave you and hit Run.

Here's what to do:

1. Click "New Notebook"
2. Go to files in the left menu (folder icon) and upload your CSVs.
3. Paste the code ChatGPT gave you (you might need to do this as multiple code blocks)
4. Click the Play button (little triangle on the top left of each block) on each code section to run, or, click run all at the top.

That's it. In about 3 minutes, you'll see:

• The data the code is using to perform the analysis
• The charts showing analysis
• The forecast at the end

Want a step-by-step video? Christian Martinez (AI Finance Club expert) has a full walkthrough here: Christian's Google Colab Tutorial​

Step 6: Export and Analyze

This is a really useful step that most people don’t do:

1. To export your data from Google Colab, ask ChatGPT:

Once it gives you the code, paste it into a new code block in Colab and click run. This should download it as a zip of CSV files you can extract.

2. To get ChatGPT to help you analyse and improve. Upload the CSVs back to ChatGPT and ask:

This is super important, because getting your first results is just the start of seeing what machine learning can do.

To really use this for forecasting or decision-making, you'll need to:

• Try different ML techniques and compare them
• Refine your variables based on what the first model shows
• Test different timeframes and assumptions

And AI can help you do all of that. Just ask.

​
The Bottom Line

Once you stop analyzing one variable at a time, you see your whole business as one connected system.

You go from reporting what happened last month to predicting what happens next. From descriptive to predictive.

So, next time when your CEO asks "Should we increase marketing spend?" - you don't need two weeks anymore. You already know how marketing connects to sales, to support costs, to cash collection. Because you analyzed them together.

Sequential analysis made you a reporter. Pattern-based analysis makes you a strategist.

And with machine learning, you can start spotting patterns in minutes instead of months.

Your Move

So here's what I want you to do this week.

Pick one thing you analyzed last month. Revenue drivers, cost patterns, customer behavior etc. Take your data, export to CSV, and follow those 6 steps, and see what patterns show up that you missed the first time around.

Start with the preliminary analysis in ChatGPT. If that makes sense, productize it in Colab. Then bring the results back to ChatGPT to understand what they mean.

This is how you learn. Not by reading about machine learning. By running it on your actual data and seeing what it shows you.

So, are you going to keep analyzing one spreadsheet at a time, or are you ready to see how everything connects?

Give it a try and let me know.

Best,

Your AI Finance Expert,

Nicolas

P.S. - What did you think of this approach? Hit reply and let me know if you're planning to try this for your team (I read all replies).

P.P.S. - The AI Finance Accelerator launches January 2026 - 6 weeks where you will deploy real AI reporting, finance agents, and ML forecasting (not just learn about them). After training 10,000+ finance professionals, we will help you implement AI that you can run the day the cohort ends. 6-8 hours/week. Early-bird pricing for the first 100 only. If you're serious about leading finance in 2026, join the waitlist here.