Artificial intelligence can help shave off hours of time when it comes to previously manual tasks, such as scenario planning or financial modeling. But as finance employees begin to rely more heavily on AI, it’s crucial for finance employees to understand its inner workings — meaning  “the CFO needs to have a whole set of processes in place around governance and ethics” when it comes to AI, according to Ian Schnoor, executive director of the Financial Modeling Institute.

The FMI offers a global program for financial modeling accreditation, a discipline where AI is becoming more entrenched. At the moment, many companies are still in the early days of adopting the technology, and “do not have strong disclosures or guidelines yet around AI,” Schnoor told CFO Dive.  “But if I was a CFO, I would want a policy to know exactly, how was AI used in this process?”

AI ‘time zero’

Schnoor, who began his career as an investment banker at Citibank, has served as executive director for the Toronto, Canda-based institute since 2016, according to his LinkedIn profile.

He also serves as an adjunct professor for the Smith School of Business at Queen’s University in Toronto, Canada. Before FMI, he served as founder and president for the Marquee Group, a financial modeling training provider he sold to Training the Street in March of 2023.

For Schnoor, when it comes to financial modeling, “time zero” — or the period where AI started to show the legitimate capacity to build models — occurred around February. This spurred a short phase where many in the industry expressed concern or nervousness that their skillset was about to become obsolete.

However, a few months following, there was a “whiplash” where the industry began to realize that “people probably need stronger modeling skills than they ever needed before” in the age of AI, Schnoor said.

Part of that change in perspective came about because, even as AI seeps further into the day-to-day functions of employees, the technology does not inspire the trust of a human analyst. When it comes to financial modeling, for example, company leadership is still seeking for a subject matter expert that knows the ins and outs of the model: “At the end of the day, the trust and the confidence comes from a human delivering a message to another human,” Schnoor said.

Rather than replacing human financial modelers, the challenge the industry is facing now is taking the next step to use AI to help expedite the modeling process while still maintaining insight and knowledge to create the confidence that the model is accurate, Schnoor said.

Setting the framework

For CFOs, having that insight and transparency is especially critical, as they bear the ultimate responsibility — and will face the consequences — for their businesses’ financial decisions, as well as the corporate strategies that stem from those decisions.

“Ultimately, if I was a CFO, the first thing I would want is a strong framework around AI usage on my team,” he said. That includes not just having a transparent understanding of how AI was utilized to develop the financial model, Schnoor said, but an understanding of what one’s employees or modelers did during that process.

“First, tell me how much AI was used, but second, tell me what human interaction did the team do?” he said.

CFOs should also spend some time themselves experimenting with AI in the early days of its implementation, spending some time with the model “doing my own checks and doing my own stress tests,” Schnoor said.

“At least in these very early AI days, I probably need to kind of dive a little deeper than I normally would, just to make sure that nothing's going to fall through the cracks because we're still developing processes and systems, and I can't risk an error built by an AI agent,” he said.  

It’s important to ensure that human judgement keeps its place in the financial modeling process as AI usage becomes more common.

“We still need to keep humans very connected to the decision making and the inputs, the thinking,” Schnoor said. Using AI can help expedite some of the manual processes attached to models, but “we cannot allow AI to arbitrarily choose inflation rates, interest rates, cost assumptions without having massive insight and oversight into what that means,” he said.