AI changes the role of the CFO less than is said, and more than it seems. It's not about an algorithm preparing reports for controllers, replacing an entire accounting staff, or analyzing data for the CFO; it's about the CFO deciding whether the investment in AI will pay off, who is responsible for it, and whether it will disrupt controls within the company. It's not IT, but the CFO, who should decide whether the investment in AI will pay off and if the company can afford it.
- Why is the decision to invest in AI more difficult for a CFO than a regular investment?
- Three questions a CFO should ask before implementing AI
- Which AI risks land on the CFO's desk?
- Note for PE funds and operating partners
- Practical framework: how a CFO should decide "in or out"
- Summary
Why is the decision to invest in AI more difficult for a CFO than a regular investment?
Investing in AI-powered tools still involves many financial unknowns. I usually take a firm stance that everything can be calculated, that ROI is ROI, etc. If so, why have as many as 74% of companies not yet shown tangible value from AI implementation? There are several reasons, starting with the fact that few companies even bother to calculate ROI, and many ideas "die" at the conceptual or pilot stage. This uncertainty, high unpredictability, and technological volatility are some of the reasons why the decision to invest in an AI tool, based on financial data, is so difficult.
Another reason is underestimated costs. While license costs are generally known, the total cost is difficult to calculate due to potential complications in system integration, testing, model retraining, and managing the risks that artificial intelligence creates. Financial analysis is sometimes merciless. In theory, digital transformation should be profitable. But when a CFO realistically looks at the costs, it turns out that the project will either pay off in several years or, with current assumptions, not at all.
This is because the tangible benefits of implementing generative AI are difficult to measure at the business case stage. In this area, the CFO's capabilities are limited. How can one calculate the benefit of better, more accurate, perhaps faster decisions? A CFO (Chief Financial Officer) usually has experience with automation projects, but calculating the benefits of implementing AI-based solutions is a challenge.
Three questions a CFO should ask before implementing AI
Before the decision "let's implement AI" is made, the CFO should obtain answers to three fundamental questions. If something is inconsistent, unclear, or if there's no answer or evasive responses, that's already an answer in itself as to whether the organization is truly ready for new financial challenges.
Do we have well-defined and tested assumptions necessary to calculate the return on investment?
Artificial intelligence is developing at a rapid pace. Costs, technology, and potential benefits are changing. What was new a year ago may be a standard today. So, is financial data analysis possible in such a reality? A modern CFO is aware of this. They know that strategic decisions involve an element of uncertainty. But this complexity and dynamism cannot be an excuse for a lack of information or unreliable data.
Who is responsible for the project?
IT, finance, the department where the change is being implemented, the board? The strategic partner of the CFO, whoever he (she) is, must be one designated person. Not to blame someone for failure, but to monitor progress, risks, and changing circumstances together with IT and the CFO.
How will AI tools impact company risks?
Changes in processes and new AI-based tools can both increase and mitigate risk. The implementation of AI itself and the creation of solutions may fall under the so-called AI Act, which outlines the obligations of companies and organizations related to AI. On the one hand, risks and obligations increase when considering issues related to personal data; using open models without additional safeguards carries the risk of data leakage. On the other hand, for mass applications, the accuracy provided by AI can significantly reduce risk. My favorite example is sensors detecting skin structure changes in animals. Why? Precisely to "catch" potential diseases or other concerning changes early enough. Of course, AI can make mistakes, just like humans. But apparently, AI eliminates its mistakes from future processes faster.
Which AI risks land on the CFO's desk (and whose are they, really)?
The CFO is not solely responsible for all risks, and not every risk should land "on their desk." What matters are the risks that, if they materialize, will have a significant impact on finances.
In fact, you don't need an AI implementation project to monitor risks. The first, and in my opinion, most important, is the risk associated with shadow AI. Everyone carries a phone in their pocket that can take a picture or ask Claude what it thinks about a given situation. And that situation might be a client contract, for example. "Write a response to the letter" - snap, picture attached to GPT. Because, of course, it's faster and simpler than thinking it up yourself. Such behavior alone, without any organizational changes, can cause significant damage, primarily related to data leakage. This is not a risk whose value is easy to estimate. Awareness of its existence allows for minimizing it before it materializes.
Project failure risk is another category. Can we calculate how much it might cost the company? Partially. The cost of the investment. In addition, there is reputational risk (if we promised employees or clients "the moon") or lost opportunities.
Risks related to infrastructure, tools, legal, communication... The list is impressive. The CFO is not responsible for each of these areas. But they are responsible for managing the finances that these risks impact.
Note for PE funds and operating partners
For Private Equity funds and operating partners, involvement in this topic has double significance, as it concerns not just one company, but the entire portfolio, and across three stages: at entry, during value creation, and at exit.
The entire due diligence process should be expanded to include AI due diligence. Not only legal, HR, or financial matters are assessed before an investment, but also AI: in the context of the tools used (or not) and, less obviously, what potential AI-based solutions offer for the company. Potential or threat. Investing in a promising company that lacks structured data, an information base, etc., will mean additional expenses for the fund. Another issue is how AI-based solutions will impact the product or service offered by the company. There used to be such professions as a repasser, elevator operator, or waiter. Oh, excuse me, waiters and waitresses are still around. Although more and more often I encounter a QR code on the table in restaurants instead of a menu, orders are made independently, and food is brought by a robot. Awareness of the industry, product, and service in the context of generative AI development is one of the elements of AI due diligence.
Practical framework: how a CFO should decide "in or out"
The simplest answer is: look at the numbers. Even for an investment whose return cannot be reliably calculated, the CFO's role is to estimate, consider, analyze, and predict. Before reaching this stage, the management (not just the CFO) should start by asking: is the company even ready (mature - AI maturity) to implement AI. Maturity is assessed in five dimensions: diagnosis, meaning whether we can name a specific business problem, data, people and organizational culture, governance, and technology. We are not yet talking about finances, calculating ROI, or assessing risks. We are considering whether we, as a company, an institution, are ready for change at all. Investing in AI should begin with this maturity assessment, because without organized information about the current state, failure is guaranteed. It's like buying a car for yourself and not having a driver's license.
Only when the company knows its level can it consider calculating ROI. An Excel spreadsheet for calculating return is simple: potential benefits on one side, costs on the other. The calculation itself is straightforward. The CFO's challenge is to gather information that can significantly impact the project and the organization.
The rule is simple. If the company is not mature, first tidy up the fundamentals instead of forcing AI implementation. If it is, and the value and cost are clear, proceed. If the value is real but uncertain, conduct a pilot. If anything is uncertain, do not proceed until it is clarified. No answer is also a decision, only then it is made consciously.
Summary
The role of the CFO in implementing AI-based tools and solutions is not just to calculate the return on investment, as CFOs have always done. It is about imposing discipline, an investment framework, and a broader perspective on the entire project. The CFO starts by asking about maturity, then about risks, benefits and costs, as well as responsibility and potential scenarios. Only then do they decide whether the organization can afford the project.
If decisions about AI in your company are made without this discipline, or if you need someone to introduce it without hiring a full-time CFO, check out the CFO as a service.
