Generative AI continues to make headlines, and for CFOs who want to take advantage of this fast-developing tech, “fine-tuning” the model is the next necessary step, said Glenn Hopper, director and CFO for finance and accounting solutions provider Eventus Advisory Group and author of Deep Finance: Corporate Finance in the Information Age.
That means carefully thinking not only about AI’s potential use cases in corporate or other finance functions, but about what data that model needs to understand and fill the business’s specific needs.
“If ChatGPT is a generalist…then fine tuning a model would be sending them to get their masters, to make them a domain specific expert,” Hopper said in an interview.
Data, data everywhere
Generative AI and LLMs’ role within the business world has garnered increasing attention since the launch of OpenAI’s ChatGPT tool just last November. Last week, Hopper participated in a webinar hosted by the Digital, Data, and Design Institute at Harvard on the changing role of AI in corporate finance.
He was joined on the panel by Alexandra Mousavizadeh, co-founder and CEO of Evident — a London-based research company that measures AI adoption and publishes the “Evident AI Index” — and Sanjay Srivastava, chief digital strategist at professional services firm Genpact.
While it’s important to be cautious of the hype surrounding what generative AI can do, the technology still has serious potential finance leaders need to consider, panelists agreed.
For example, one of the major changes brought about by generative AI’s development is that “the next programming language is English,” Srivastava said during the webinar — meaning that business analysts can express a question in natural language, with the AI then taking the necessary steps to convert the question to a programming language, access the database, and bring back the key data.
“What that is doing is it’s transforming the way we think about automation and business analytics,” Srivastava said.
As more and more companies try out generative AI, understanding the role and importance of the data underpinning these models is crucial for CFOs. Data is “foundational” when it comes to training AI models, Hopper said, meaning access, quality and quantity of information are all key factors for businesses at the dawn of the AI age.
“If you're a dry cleaner with two or three locations, I mean, there's a finite amount of data you have to train the model,” he said. “If you're a service provider, and you have 1,000 dry cleaner clients, you could use their data to actually train it.”
Generative AI has already triggered data security and privacy concerns, with Sam Altman, CEO of Microsoft-backed OpenAI, fielding questions on Tuesday from lawmakers during a hearing held by the Senate Judiciary subcommittee on privacy, technology and law on the regulation of generative AI.
The importance of what data is being used to train AI models also opens up questions of competition as well as data transparency — if two competing companies are using the same generative AI or LLM tool, for example, the businesses may not want their information used to train the AI their competitor is using.
Furthermore, not “everybody has the luxury of having as much data as Amazon, Facebook and Google and all the other big tech cloud providers,” Hopper said. “So they can’t capitalize on this.”
AI will shape the future CFO
As experts and lawmakers debate AI’s regulation, keeping pace with the technology’s swift advancement is a must for CFOs who don’t want to be left behind the curve. While finance chiefs don’t “all have to be engineers,” it’s critical for financial leaders to “understand the fundamentals of (the tools) before we start incorporating them into our workflow,” Hopper said.
This is especially critical as generative AI may not just impact the way CFOs approach business operations, but the future of their role itself. Looking to the future, Hopper is hard-pressed to find a job that AI could not potentially replace, he said last week during the D^3 webinar — including the role of the CFO.
“It's hard to be in that role and answer in the affirmative,” Hopper said, when asked if he could see a future where AI does overtake the CFO position. “But I think we have to act as if that is a possibility and not bury our heads in the sand.”
Whatever the case, it’s clear the CFO role in the future “won’t look like it does now,” given these advancements in technology, he said. However, Hopper also pointed out that “the CFO role now doesn't look like it did 20 years ago.”
“The biggest thing for us is, we have to be aware of and understand the technological environment that you're in and evolve with it and keep finding ways to work with the technology to keep adding value,” Hopper said.