Hi Reader,

When my boss asked me to create a 5-year scenario plan in Excel, I used to open a blank workbook, stare at it, and tell myself, “This financial model should only take me an hour.”

6 hours later (often late in the evening), I was still wiring tabs and chasing circular refs.

But now this can be done in 10 minutes with AI.

Today I am going to show you my playbook to build a 5-year strategic plan with three scenarios in less than 10 minutes (+ a way to show it to your boss in a visual way to impress them).

Before diving into today’s topic, I have an important update to share with you first…

Tell The Right Story with Your Data

Join me next week for a free masterclass on using AI to clean data, sharpen analysis, and communicate clear financial stories that drive decisions (and show you how to use it for financial modeling).

​The link to register for free​

Here's what you will learn as well:

• Why financial storytelling is critical for modern Finance teams
• How AI automates data prep and improves trust in numbers
• Smarter ways to analyze and present insights with AI support
• Practical tips for creating and sharing impactful reports

This event is perfect for CFOs and Finance Teams who want to save time and conduct better and more precise analysis of their numbers

This newsletter is sponsored by Jedox

What “Financial modeling with AI” actually means and how to do it

Now that GenerativeAI has the reasoning mode available, we can build, update, and explain finance models faster and with tighter controls.

A reasoning model is a model which works task by task and think about the goal first, and then will iterate on all the tasks needed before giving you the answer.

The AI would be like an assistant that would draft structure, link formulas, run scenarios, check reconciliations, and generate narrative.

How to use it?

Imagine your boss comes to you and wants to have a comparison of 3 different scenarios over the next 3 years:

• Base scenario with normal growth included factored in
• US Expansion: adding additional sales from the US
• Pessimistic: with a struggling economy

Below is how I prompted AI:

Give the baseline

• I provided AI with the 2025 P&L to have both our templates and our actual figures inside

Be super precise on the assumptions you want for each scenario:

Here you can see that I am stating the increase I want and the frequency:

• Base: revenue +3% YoY, margin constant, costs +2% YoY
• US Expansion: adds a gradual US ramp (incremental revenue and related costs, typically controlled on the assumptions sheet)
• Pessimistic: sales volume −2% YoY, margin −1pp per year, other costs - 1% YoY (i.e., only a small improvement in costs)

Other tips from my experience after running dozens of models with AI:

Provide minimum specifics per scenario: at least set revenue growth, margin behavior, and cost growth/one-offs.

If you leave these undefined, the AI will make its own assumptions, and the result may be different from what you want.

Supply a baseline P&L template. Give the AI your P&L structure as the base so it keeps the same line-item layout across all scenarios.

This forces structural consistency and makes comparisons and roll-ups straightforward.

Ask for formulas, not just values. Specify that outputs must include the underlying Excel/Sheets formulas and that every scenario references the Assumptions sheet drivers.

This ensures the model remains fully dynamic and auditable.

Here you see the prompt and the result:

Here's the prompt I used:

​The link to the file used in the example​

Below is a screenshot of a sample Excel sheet output from ChatGPT.

It shows an Assumptions tab that drives a three-scenario P&L model with tabs for Assumption, Proj_Base, Proj_US_Expansion, Proj_Pessimistic, plus a Readme.

Overall, the structure is clean with a single source of truth for drivers, clearly separated scenarios, and consistent P&L outputs - ideal for CFOs who want fast scenario switches without rewriting formulas

Really impressive, right?

Yours may not be identical, as generative AI is non-deterministic, so two runs of the same prompt can vary.

To increase consistency, you can be more specific and structured with your prompt. You can specify more sections, headers, formulas, and formatting that you expect. Clear and concrete instructions lead to more repeatable results.

Bonus: how to visualize your financial model (to impress your boss?

Now, how to visualize your financial model?

In order to do this in ChatGPT, enable a hidden function which only a few people know: Canvas.
​
You might ask: What is Canvas?

Canvas is a side-by-side workspace where you and the model co-create charts, tables, and slides in one place. No need for context-switching between chat and docs.

How to use ChatGPT Canvas?

1. Open Canvas: type “/canvas” (or click the Canvas toggle if you see it).
2. Provide data: paste a small table or upload a CSV/XLSX. State units (currency; thousands/millions) and time grain (M/Q/Y).
3. Ask for the visual: specify the chart type + dimensions.
   • Example: “Build a variance waterfall for 2025 vs 2024 from this table. Bars: Price, Volume, Mix, FX, Cost. Show both $ and pp impact. Add labels and a total column.”
4. Refine: request titles, axis labels, rounding, color-blind-safe palette, and annotations (e.g., “flag margin < 20%”).
5. Export: copy the chart image into your deck or brief. Keep the data + prompt under the chart for the audit trail.

Good finance visuals to request

• Variance waterfall: Actual vs Plan (or vs LY) with drivers (price/volume/mix/cost/FX)
• Bridge charts: Revenue - EBITDA - FCF
• Cohort heatmap: customer revenue by start quarter (retention/expansion)
• Small multiples: GM% by product line across quarters
• Working-capital dashboard: DSO, DIO, DPO, CCC trend with target bands

Look at the GIF (until the end):

Activate Canvas and use the prompt I used in the example:

The quality checks you need to always do:

I start by checking the labels: every chart shows units and rounding, and totals tie back to the source table.

Then I look at the axes. They begin at logical baselines, so nothing is exaggerated.

I add clear threshold callouts - for example, “GM% below 22%.”

I include a simple footnote with the data source and the period it covers.

Finally, I iterate: rerun the prompt, check clarity and accuracy, and lock the constraints (units, rounding, axis min/max, threshold labels) until the visual reads clean and reliable.

This is how you impress your boss!

Why this matters

I keep repeating this in my AI CFO sessions: scenario modeling + automation is becoming table stakes for finance leaders. Imagine the hours saved by applying this simple technique

If you can build, switch, visualize, and narrate scenarios in minutes, you are already many steps ahead of your peers, and you become invaluable to your boss.
​

Best,
Your AI Finance Expert,
Nicolas

P.S. My latest video tutorial on building financial models is out and has already been watched by more than 5,000 finance pros just like you.
​Here's the link if you want to check it out.​

Transparency note:​
Some of the links in this newsletter are affiliate links or paid placements. That means I might earn a commission if you use them. But I only share things I personally trust and believe are worth your time.