72 CFOs budgeted FY2026 in 30-mins with this 1 method

15 years’ CPA experience, and then this:

"Demonstrated experience using generative AI tools."

5 years ago, that line wasn't on any job description. Today it's on the highest-paying ones. In 2 years it'll be on all of them.

So, to make sure you don’t fall behind in your career, join me for this free 60-min masterclass. I'll show you:

• How to build a 5-year, 3-scenario model in minutes (and use it in a board meeting as a live tool)
• How to auto-generate the branded PowerPoint from the same model (without formatting a single slide)
• The one Copilot setting 90% of finance pros don't know exists

1,332 pros joined me last time. 95% stayed for the full hour.

So, to get ahead of AI and improve your career, make sure to save your seat here before we are full.

Only 15% of CFOs have scaled AI in finance (my members are proving them wrong)

Bain just surveyed 102 CFOs globally.

The crazy number: only 15-25% of CFOs have scaled AI across finance.

But, the same survey said something else super important. The biggest payoff so far with AI is not cost reduction, it’s speed.

This sounds great in a Bain report, but this is just theory. I am here to give you something practical.

So, a few weeks ago we ran the second MBA case study session inside the AI Finance Club.

We gave our members a made-up construction company. It had messy raw data with multiple tabs and missing figures.

Then we told them build an FY2026 budget the lenders would sign off (in 30 minutes)

One of the participants, Marcin Rudnik, did it with two files and a 3-prompt stack. He created the board pack, the dashboard, and the covenant analysis in just one short session.

Today I'll show you the method he used.

​

​

“I need an FY2026 budget the board will sign off on next week”

The method we teach? ‘Prompt Stacking’.

This is where you take multiple prompt frameworks, and use them together to achieve a specific result.

Here's a quick breakdown of the frameworks we taught in our workshop.

(you’ll see Marcin’s twist a little later)

CSI: Context, Specifics, Instruction.

Before diving straight into data you need a brief.

Sol in your prompt. Define the role, name the problem, and give a clear instruction.

That's exactly what Marcin did. And Claude came back with three options: lender-first covenant-driven, project-by-project bottom-up, and AI-native interactive dashboard.

He picked the third because of the time pressure.

So. With CSI, your brief is done.

Now your data has to make sense to AI – because if it doesn't, AI guesses.

This is where our next framework comes in.

SNAKE: Specify, Note, Action, Key data structure, Extra details.

This one is for handing data to AI when you want it to write Python code for you. The framework is called SNAKE because Python is a snake. (Christian's joke, not mine).

• Specify the file type ("My data is in an Excel file called SummitBuild_Case_Data_FY2026.xlsx").
• Note the environment ("I'm using Google Colab. Prompt me to upload the file").
• Action ("Generate Python code to build a monthly cash flow dashboard with variance analysis vs. prior year").
• Key data structure (paste the headers plus five rows: "Month | Revenue | Direct Costs | Gross Profit | EBITDA | Operating Cash Flow").
• Extra details ("Use a dark theme. Include a summary table and a bar chart. Flag months where operating cash flow is negative").

The Key data structure step is the one that stops hallucinations. AI stops guessing what your data looks like, because you told it.

Note – You can use the same framework even if you don't want to use Python. 'Note the environment' could be "Excel". 'Action' could be "Monthly cashflow model using VBA / Office Script / Google AppScript" etc. The point here is helping AI take the guesswork out of your data, and reducing room for error.

Now your data is briefed. Python (or another tool) is ready to run.

Now you need to tell AI what the output has to look like (plus who's reading it).

FBI: Format, Blueprint, Identity.

This is for the deliverables.

1. Format ("interactive HTML dashboard").
2. Blueprint ("flag months where operating cash flow is negative; include base, downside, upside").
3. Identity ("write commentary as a fractional CFO presenting to a covenant-watching lender").

And that's it.

Each framework on its own is fine. But stacked, they replace 80% of the back-and-forth that ruins your evenings.

There're a couple of other important steps after this that you'll see in the 30-Day-Challenge below.

But first let's have a quick look at what Marcin produced.

​

Marcin’s Twist

This is the part that surprised me.

He uploaded the two files. But then instead of just copy pasting the prompts, he asked Claude for the best prompt based on our frameworks (super efficient).

What came back was an interactive HTML dashboard he could manipulate live: pipeline win rate, labor wage inflation, material cost inflation, DSO, G&A growth. All adjustable in the UI (User Interface). The numbers on the P&L and the cash flow updated in real time.

Then he asked for the board presentation. Claude came back with 13 projects analysed, gross pipeline value, probability-weighted revenue, contract signed status, backlog coverage, a P&L summary with growth and margin expansion commentary, and an internal covenant compliance check.

That is more than most teams produce in a week.

He had it inside 30 minutes.

But here’s something super important.

When Marcin presented, he didn't pretend it was magic. He talked about what was missing.

There was no balance sheet. No comparison between historical cash flow and planned capex. Plus, the 100% pipeline win rate Claude pulled from the assumptions was, in his words, "crazy." So he manually adjusted it before sensitivity testing.

His exact comments:

This is always the question you should be asking.

AI is the junior who builds fast. You are the CFO who decides what to trust.

Which is why the next step after any AI-built model is the data audit.

I broke down exactly how to audit your data for accuracy in my previous newsletter 'Steal my AI accuracy method (from 7 years in audit at PwC)' which you can read here.​

​

​

The 30-Minute Challenge

So. Here's my challenge to you. Pick one upcoming deliverable. A budget, a forecast refresh, a board pack. Use this process to build it before next quarter.

Step 1. Open with CSI (before you touch the spreadsheet).

Sit down before you open Excel. Write three lines.

1. Context: who you are, what the company does.
2. Specifics: the deliverable, the deadline, the audience.
3. Instruction: what you want AI to produce first. Usually three competing approaches, not the final model.

Step 2. Brief the data with SNAKE.

Specify the filename. Note the tool (Google Colab, Excel etc). State the action in one line. Give Key data structure: paste the headers and five rows. Add Extra details: theme, charts, flags you want raised.

This is the step that stops hallucinations.

Step 3. Produce the deliverables with FBI.

1. Format names the artefact (HTML dashboard, board-deck section, lender memo).
2. Blueprint names what has to be inside (variance flags, three scenarios, covenant ratios).
3. Identity names the voice (fractional CFO, audit partner, treasurer).

Step 4. Run sensitivities.

Once the dashboard exists, change one assumption at a time and watch out for what breaks.

This is the step that turns a model into a board conversation, because you stop saying "here's the budget" and start saying "here's what has to be true for this budget to work."

Step 5. Save as a template.

The whole point is repeatability. Save the CSI brief, the SNAKE data block, the FBI deliverable spec into a single document.

Next quarter, all you need to do is change the content.

​

​

The One Thing to Remember

Everything becomes easier once you have a framework.

Here you have a 3-prompt stack that you can copy and start using straight away.

Just don’t forget that what gave Marcin his edge during the workshop. It was his ability to think critically, and use his brain space and gut feel to call out something that didn’t look right.

Speed is the easy part. AI gives you that.

The hard part is knowing where to push back, and which assumptions to refuse before you present, as this is what earns you credibility.

Your prompts do the building. You do the thinking.

Best,

Your AI Finance Expert,

– Nicolas

​

P.S. – Will you use this method? Hit reply and let me know (I read all replies)

P.P.S. – I made a video showing exactly how I build a live financial dashboard with AI. You can watch it here and then do it yourself straight away.

​

​link​