So, you've spent the morning building your forecast. The trend line is clean, the growth rate looks sensible, and you're feeling pretty good about it.

You show it to your colleagues, and within ten seconds they look at you and say:

"But our Q4 numbers always spike at year end. Why haven't you taken this into consideration?"

And just like that, your forecast goes from "done" to "do it again."

When you prepare a forecast, most of the time you look at historical data, but if there is seasonality in your business you want to reflect that.

If March is a good month because nobody's on holidays and there are 31 days, you want to show higher sales. If August is lower because people are on summer holidays and you have less activity, you want to show that too.

As humans, we do that with some formulas or we do some plugs in our forecast, but it's not really accurate and it takes a LOT of effort.

But, fixing this with AI is much simpler than you think. You can do it in about 30 minutes, and you don't need a data science degree to get there.

​
Flat-Line

If you just ask AI to produce a forecast for you, it might not be bad, but probably will not show the full picture.

A linear trend will capture the general direction of your business. Revenue is growing at roughly 8% a year, great. But it doesn't know that January is always slow and December always spikes. It doesn't know your business runs in cycles. So it draws a straight line through a world that moves in waves.

If we ignore these predictable cycles, our forecast will overestimate revenue in the slow months and underestimate it in the peak months.

And this is where a lot of people get stuck, because AI knows how to create code, but it doesn't know your business. It doesn't have your business knowledge or know the business impact.

One of our AI Finance Accelerator participants, Jürgen, ran into exactly this. He built his first forecasting model and the output was a flat line. It was going upward, sure, but it wasn't reflecting reality.

So he iterated.

He told the AI about the seasonality in his data. He accounted for a COVID effect that had changed specific periods. And, after a few rounds of back and forth (where he brought his business knowledge and the AI brought the code), he ended up with a result that was really good.

That iterative process is what matters. You are the one who knows the business. The AI is the one who writes the code. Together, you create amazing results.

​
Seasonal Superpowers

So how do you fix a flat-line forecast? With a single multiplication.

You take your trend forecast and multiply each month by what's called a seasonal index. This is just a number that tells you how much a given month typically changes from the annual average.

An index of 1 means that month is exactly at the annual average. If December has an index of 1.3, that means it's typically 30% above the annual average. If January has an index of 0.85, it means it's 15% below.

So, you take the trend value, multiply it by the seasonal index, and you get your seasonally adjusted forecast.

Two components (trend and seasonal index) and your forecast now captures where your business is heading plus, how it moves through the year. It's still simple, transparent, and auditable.

​
How to Do It in 4 Steps (30 Minutes)

You can do this with any AI (just make sure it’s a secure version), and run the code in Google Colab (so you don’t need to install anything).

For a deeper dive on using Python for forecasting, check out my previous newsletter here.

For now, we’re just looking at seasonality.

Step 1: Start with your trend forecast. If you already have a basic linear forecast (even built in Excel), export those monthly projected values as a CSV. If you don't have one yet, ask your AI tool to generate a simple linear regression forecast from your historical data. That becomes your baseline.

Step 2: Calculate the seasonal indices. Upload your historical monthly data and prompt your AI with something like:

The AI will generate the Python code for this, and within seconds you'll have your indices. You don't need to write or understand the code yourself. You just need to understand what the output means.

Step 3: Apply the indices to your forecast. Prompt:

When you see those two lines side by side, you'll immediately notice the difference. One is a straight line that goes up. The other has the rhythm of your business in it.

Step 4: Validate with what you know. This is where your finance knowledge becomes the most valuable part of the whole process. Look at the output. Does December spike the way you'd expect? Is Q1 softer? If something doesn't match what you know about the business, tell the AI. Say something like:

And adjust. The AI wrote the code, but you are the one who decides if the output reflects reality.

One thing to keep in mind: if you only have two years of historical data, your seasonal indices will be noisy. Three or more years gives you much more stable patterns. Tell the AI how many years you have so it can tell you about any reliability concerns.

The One Thing to Remember

Your colleagues did not reject your forecast because the math was wrong. They rejected it because it didn't look like the business they see every quarter.

Plus, if you document how you built the indices and where the data came from. You're showing how you worked it out, and that builds trust.

One multiplication (trend times seasonal index) is the difference between a forecast that gets sent back and one that gets discussed in the next leadership meeting.

It takes 30 minutes. And next time your colleague asks where the Q4 spike is, it'll already be there.

Best,

Your AI Finance Expert,

Nicolas

P.S. - Have you tried adding seasonality to your forecast before? Hit reply and let me know how it went (I read all replies).

P.P.S. - For those of you using Gemini (or thinking about it): I put together a guide that goes way beyond basic chat. We cover how to analyze data with the latest model, visualize results in HTML, and connect deeper with your Google ecosystem including Google Colab that I think will change how you use it. Watch it here → How to Use Gemini For Finance (Full Guide 2026)​

​