For deterministic problems, use deterministic solutions.
The current hype around Artificial Intelligence has convinced many people that AI should be applied everywhere — from writing marketing copy to coding apps, from designing logos to, also, building financial models. But let’s take a step back. AI is amazing for brainstorming, idea generation, pattern recognition, and even reasoning about messy, uncertain problems. You can use it to explore business assumptions, challenge your thinking, and analyze large datasets. AI excels at tasks where creativity, iteration, and probabilistic reasoning are useful.
Financial modeling, however, is a different story, because its process is deterministic in many ways.
Financial Models Are Deterministic
A financial model is not an open-ended creative exercise. It is a structure built on well-defined financial and economic rules.
- Depreciation: Straight-line depreciation of a $100,000 machine over 5 years at 20% per year is not open to interpretation. There is one correct schedule.
- Loan amortization: The interest portion vs. principal repayment is defined by the amortization formula. An AI might invent alternative splits — but that’s wrong.
- Balance sheet consistency: Assets must equal liabilities + equity. Always.
These rules are deterministic — they follow formulas, accounting standards, and logical consistency. An AI model that “hallucinates” even slightly can compromise the integrity of the entire financial framework. And unlike creative writing or image generation, even a small error in a financial model can have serious business consequences.
Think of an investor presentation where the cash flow statement doesn’t balance because an AI-generated formula missed a step. That’s not innovation — that’s reputational damage.
The Risk of Losing Control
Financial models are already complex and error-prone: 88% of spreadsheets contain errors according to PwC, or is it 94% (according to phys.org)?
I’m not sure, but imagine building that model with AI “help.” You risk:
- Harder-to-detect mistakes, like a shortcut formula in Excel that breaks circular references without the modeler noticing.
- Likewise, the loss of logic traceability matters. Example: a revenue growth formula that multiplies by the wrong base year — it looks correct at first glance, but derails projections.
- Compliance pitfalls, perhaps by adopting the wrong context. Suppose AI inventing a “creative” tax calculation that doesn’t comply with local regulation.
One finance director friend of mine has put it bluntly: “If I don’t know how the number was calculated, I can’t trust it.” That’s the heart of the issue. In financial modeling, control and transparency are non-negotiable.
The Value Is in the (Thinking) Process
One of the greatest benefits of building a financial model is not the final Excel file — it’s the learning and discussion that happens during the process.
- Understanding causality. Think: discovering how a 5-day increase in Days Sales Outstanding (DSO) can wipe out $200,000 in cash flow.
- Sharper capital decisions. Think: finance proposes funding growth with more debt, while strategy pushes for equity financing — the debate reveals the trade-offs between leverage risk and shareholder dilution.
- Operational visibility. Think: the exercise of linking sales volume growth to headcount needs reveals operational bottlenecks early.
The model is not just a tool for forecasting. It’s a training ground for financial thinking and team alignment.
If the modeler does not think, he does not learn. If he does not understand what is being built, neither will his teammates.
Learning Is Human
Learning is not something to be outsourced to AI. It’s human, messy, and iterative. It thrives on criticism, questioning, and dialogue.
- In a project finance model, debating the assumptions about energy prices teaches the team more than the spreadsheet itself.
- In a startup financial plan, the discussion around churn rates and customer acquisition costs helps align marketing, sales, and finance on strategy.
By outsourcing the build to AI, you don’t just risk errors. You lose the very discussions that sharpen your assumptions, surface blind spots, and build confidence among stakeholders.
When to Use AI in Financial Modeling
So when should you use AI? Not to build the model — but to make sense of it.
- Use AI to validate your idea.
- Let AI analyze the results and summarize them in a way that is easy to understand.
- Use AI to generate new ideas and scenarios you may not have considered.
- Compare your model with what is going on in the world: does it make sense? Is your idea valid? Is it innovative?