Most finance pros I've spoken to in the last year have said some version of the same thing to me.
​
"I know I should be doing more with Python and forecasting models. But I don't even know where to begin."

And then they go back to their Excel file and carry on.

What I kept hearing from people in our live workshops was exactly this: "I'm scared of Python. Do I have to learn it? Do I need to be professional in it?"

The answer to this is no.

You can use AI like a highly capable analyst who happens to write perfect Python.

Your financial knowledge is still the skill. But you do need to know the principles of using AI to generate Python to get the best (accurate) results.

Last Thursday I covered WHEN to choose Excel and when to choose Python for forecasting (read it here).

Today I'm showing you HOW to to use use Python for forecasting (even if you've never written a line of code in your life).

You'll see how to build a 6-month revenue forecast using AI and Python, from scratch, in under 60 minutes - using the SNAKE method.

​
The problem with Excel isn't Excel

I want to be super clear about something, because I'm not here to tell you Excel is broken. It isn't. For simple forecasting, small datasets, or for a CFO who knows the model inside out, it works.

The problem is what happens when your business outgrows it. You need to run the same forecast across 50 product lines.

You need to handle seasonality properly, or model multiple scenarios at once. And at that point, Excel isn't broken, it's just not built for the job.

The reason Python feels like the scary alternative is because we've been told it requires a completely different skill set.

That you need to become a data scientist before you can use data science tools.

This is just not true anymore.

​
Your Python ‘SNAKE’ Brief

Python is a programming language that can process, analyze, and model data at a scale and speed that Excel can't match.

You don't learn Python. You learn how to give AI a precise financial brief.

You tell it what data you have, what you're forecasting, and what your business cycles look like.

The AI writes the code, and then you run it in Google Colab (which is free and requires zero installation) it runs right in your browser. You then get a working forecast with a chart.

The more specific your brief, the better the output, exactly like managing any team member.

This is where you will use SNAKE:

1. Specify - Tell the AI the file type you're working with (e.g. CSV, Excel)
2. Notebook - Note the notebook environment you're using (Google Colab - which is what's called a Jupyter Notebook - a tool specifically used run code in machine learning and analysis projects)
3. Action - Tell the AI the action you want it to perform
4. Key data structure - Share your headers and around 5 rows of sample data so the AI understands what it's working with
5. Extra details - Add any business context that will affect the output (seasonality, one-off events, how you want results exported)

Plus, if the output isn't right, you don't debug the code yourself either, you tell the AI what's wrong and ask it to fix it.

​
How to build your first forecast in 5 steps

Here is the exact process. It is simpler than you think.

Step 1 - Prepare your data brief

Export at least 24 months of monthly revenue data to a CSV file. Before you open any AI tool, note the business context that the AI won't know: which quarters are typically strong or slow, any one-off events that changed your numbers (a big contract win, a COVID dip, a product launch), and whether you have missing values you'll need to handle.
​
Also note down your column headers and find five sample rows. You'll paste these into your prompt so the AI understands exactly what it's working with later.

Step 2 - Open your tools

1. Open Claude or ChatGPT Business, Gemini or Copilot

2. Go to Google Colab (colab.research.google.com) and open a new notebook.

Step 3 - Prompt using the SNAKE method

Structure your prompt around five elements:

• Specify (I'm attaching a CSV file with 24 months of monthly revenue data)
• Notebook (please write code to run in Google Colab)
• Action (clean any missing values)
• Key data structure (my headers are...)
• Extra details (Q4 is always our strongest quarter due to year-end budget cycles)

Here's an example of what that full SNAKE prompt looks like in practice:

Step 4 - Run, look, and iterate

Paste the code into Google Colab and run it.

Look at the output chart and ask yourself whether it looks like your business. If the forecast is a flat line going forward, it means the model is ignoring your seasonal patterns (the Q4 spike, the summer slowdown, whatever is true for you).

If this happens, just tell the AI:

It will rewrite the code and you can run it again.

If there are gaps in your data (months where numbers are missing), tell the AI how you want to fill them. The simplest instruction is:

Each back-and-forth with the AI takes about 30 seconds, so don't be afraid to iterate.

Step 5 - Export and present

One of the most common questions in our workshops is about auditability.

The answer is simple: at the end of your prompt, ask the AI to save your results dataframe as an Excel file.

Then you can download it, share it, and management can see the numbers without needing to touch any code.

A Note on Algorithms

A linear regression model often gets you 80% of the way there, and it's defensible. If you walked into a board meeting with a forecast built on a model nobody can explain (where the numbers come out but nobody knows why) the first question would be "why is it predicting that?" and you'd have no answer.

So, start simple, and before you present anything, sense-check it against data you already know. Take your first two years of data, ask the AI to forecast year three using only that, and then compare the prediction against what actually happened in year three. If it's broadly right, you can trust it. If it's way off, tell the AI what it missed and iterate.

Once that basic model is working well, you only need to go further if it's clearly not capturing something important about your business.

For example, if you have very strong seasonal swings that the simple trend line keeps getting wrong. At that point, algorithms like Prophet (developed by Meta specifically for business forecasting) can handle that complexity for you automatically, and you'd still be generating the code the same way via AI.

Pro tip - If you’re curious to learn more, ask AI to explain the different Python forecasting algorithms, what they work best for, and how they work.

The One Thing to Remember

There's an analogy that stuck with me from one of our AI Finance Experts Christophe:

"It's kind of like comparing a shovel with an excavator. Both will do the job. But the other one is doing it much faster in the evening instead of in a week."

You were never missing a Python skill. You were missing a workflow.

So, open Google Colab this week and try it on your last 24 months of revenue data. You won't write a single line of code yourself, and it will take less time than your last Excel version.

The finance professionals who are staying ahead right now are not the ones learning to code. They're the ones learning to brief AI in the best possible way.

Your domain knowledge is the advantage (it always will be)

And now you know how to show it off better ;)

Best,

Your AI Finance Expert,

Nicolas

P.S. Have you tried this already, or are you still on the fence about Python? Hit reply and tell me where you're at. I read every reply.

P.P.S. If you want to go deeper on Python in Finance, we cover all of this step by step inside the AI Finance Accelerator - complete this form to join the waitlist for our exclusive March Cohort.