Module 520 min read · AI in Finance

Building Financial Models with AI

Financial modeling — projecting a company's future financials to inform a decision or valuation — is where finance professionals spend enormous time. AI can accelerate the structure, logic, and even the formulas of a model, but here more than anywhere the Module 2 lesson is law: the actual computation belongs in a spreadsheet, not in the language model's predicted text. This module shows you how to use AI to build better models faster without letting it fabricate your numbers.

What a financial model actually is

A financial model is a structured set of assumptions and the calculations that flow from them, usually built in a spreadsheet, projecting future financial outcomes. The most common is the three-statement model — projecting income statement, balance sheet, and cash flow together so they remain internally consistent. From there you get derived models: discounted cash flow (DCF) for valuation, scenario models, budgeting models, and so on.

The core truth of modeling: a model is only as good as its assumptions. The math is the easy part. The judgment about what assumptions to make — growth rates, margins, capital needs — is the entire game. This is exactly the split that determines how to use AI.

The division that makes AI safe for modeling

Use AI for the structure and logic of the model — what line items to include, how they relate, what drives what, what formulas to use. Keep the actual calculation in a spreadsheet (Excel, Google Sheets) or in executed code. Never let the language model be the calculator for a model that informs a real decision.

How AI accelerates modeling

Designing the model structure
"I'm building a three-statement model for a SaaS company. What line items and drivers should I include, and how should the three statements link?" AI gives you a professional structure instantly, including links you might forget.
Writing the formulas
"What Excel formula calculates the ending cash balance that flows from the cash flow statement to the balance sheet?" AI is excellent at producing correct spreadsheet formulas you then place in your own model.
Pressure-testing assumptions
"My model assumes 25% annual revenue growth for five years. Is that reasonable for a company at this stage in this industry? What would make it too aggressive?" AI as a sounding board for assumption realism is genuinely valuable.
Explaining and documenting
"Write a clear explanation of the assumptions and methodology in this model for someone reviewing it." AI turns your model into a documented, defensible piece of work.

The spreadsheet-AI workflow

The most robust modeling workflow keeps each tool in its lane. Tools like Claude for Excel and Gemini in Google Sheets are bridging the gap — they can operate on your actual spreadsheet — but the principle holds: real numbers live in the sheet.

StepToolWhat happens
1. Design structureClaude / ChatGPTAI proposes line items, drivers, and statement links
2. Build in spreadsheetExcel / Google SheetsYou construct the model; real formulas compute real numbers
3. Generate formulasAI assist in-sheetAI writes specific formulas you verify and place
4. Run scenariosSpreadsheetChange assumptions; the sheet recomputes deterministically
5. Interpret & write upClaudeAI helps explain what the outputs mean
The cardinal modeling sin

Asking an LLM to "build me a 5-year DCF for this company" and accepting the numbers it types out. It will produce a beautifully formatted model with completely fabricated intermediate calculations. The output looks like expert work and may be arithmetically wrong throughout. If the model matters, the math happens in a spreadsheet or executed code — full stop.

Sensitivity and scenario thinking

The most valuable thing a model does is show how outcomes change as assumptions change. AI is excellent at helping you think through which variables matter most and what scenarios to test.

"My valuation is most sensitive to the discount rate and the terminal growth rate. Help me design a sensitivity table and reasonable bull, base, and bear scenarios for each." Then you build and run that sensitivity in the spreadsheet, where the recalculation is real.

A model is an argument, not an answer

The point of a model isn't the single number it spits out — it's the structured thinking about what drives the business and how outcomes range under different assumptions. AI helps you build that thinking faster and document it better. But the assumptions, and the judgment about which scenario is most likely, are yours. A model with someone else's assumptions is someone else's model.

Next

Module 6 focuses on one of the highest-value AI applications in all of finance: rapidly analyzing earnings reports, SEC filings, and other long financial documents — turning hours of reading into minutes of focused insight.