CFOs are investing in generative AI and predictive AI solutions for functions other than the one they lead, their hesitance born of worries concerning the accuracy of financial reports.
By Russ Banham
Chief Executive’s StrategicCFO360
At the large Dallas, Texas-based general contractor Rogers-O’Brien Construction, CFO Austin Robertson sees great potential in the use of AI solutions in the finance organization he heads up. Someday.
“We’re putting more and more historical data into predictive analytics solutions to forecast where we may overspend on future projects,” says Robertson. “That isn’t predictive AI, but it’s where we’re leaning. It’s not necessary for us right now to be on the bleeding edge.”
Robertson is far from alone among finance chiefs interested in predictive AI and generative AI but leery about jumping in feet first. CFO David Damiani at wealth management advisory firm Balentine says his investments have been minimal, with team members using three generative AI solutions—Zoom AI, Adobe Firefly AI and Microsoft Copilot, primarily to write emails and summarize conversations and meeting minutes.
“I’m very interested in using predictive AI in finance, but I’m concerned about the AI hallucinations,” Damiani says, in reference to incorrect or misleading outputs generated by the models, just one of several shortcomings of concern to many CFOs.
Donna Ebeling, CFO at IT consulting firm York Solutions, says her finance teams also are “dabbling” with generative AI solutions. “My interest in bundling together AI tools to step up my organization to model the numbers for forecasting purposes is extremely high, but we’re in a holding pattern now to see what works.”
While many CFOs are funding investments in AI solutions across HR, marketing, legal, IT, procurement and other functions, they’re restrained when it comes to their own organization. Nearly two-thirds (61 percent) of finance chiefs in a 2023 survey by Gartner either have no plans to implement AI in the finance function or are in the initial planning stages.
This appears to be the case at Cisco Systems, whose CFO Scott Herren recently told The Wall Street Journal about his interest in using generative AI in the finance function but acknowledges his hesitancy in moving forward. “We’re not there yet,” Herren says. “I think we need to define it well enough and then understand both the cost and the benefit before we had too far down that path.”
Aside from their difficulties assessing the return on an investment in AI for the finance organization, other factors are hindering CFOs from moving forward:
- The pressure to modernize, compelling the wrong choices. Many CFOs are just beginning to understand the differences between predictive analytics and predictive AI, much less how these tools can drive finance team productivity and efficiency gains. The very few AI use cases within finance organizations are not a confidence booster.
- Data accuracy, security and privacy. CFOs are concerned about finance teams using predictive AI tools that draw insights from sub-optimal data, as well as generative AI tools that may inadvertently expose sensitive performance data to competitors, creating financial stability risks.
- Bad outputs, the aforementioned hallucinations. AI model outcomes are as accurate as the underlying data. If data is low-quality, insufficiently trained or corrupted by biases, among other failings, and algorithms are poorly constructed, incorrect assumptions may foster bad decisions.
- Workforce needs and implications. Skilled technical talent is required in finance to develop, use and vet AI tools and solutions, talent that is extremely scarce.
- Slowness among best-of-breed finance software vendors in embedding AI in their applications. As one CFO put it, “If they’re taking their time, we should, too.”
Early Stages
StrategicCFO360 reached out to consultants at McKinsey, Deloitte and EY to expound on the impediments slowing down the adoption of these tools and solutions and, more importantly, how they can be overcome. Four finance chiefs also commented on their own AI interests and challenges.
All the interviewees see AI as a way for CFOs and the finance function to further reduce the time spent processing numbers to strategically advise colleagues across the rest of the enterprise about the meaning of the numbers in their respective functions. Using machine learning algorithms, both predictive AI and generative AI assist this transformational journey.
Predictive AI, for instance, leverages historical data to identify patterns and trends useful in forecasting future events, while generative AI produces new data from existing data across the finance function value chain to make other predictions. “CFO clients doing this well have been able to create a cash flow forecast using generative AI in hours, compared to two to three weeks previously,” says Myles Corson, Global and Americas Strategy and Markets Leader for Financial Accounting Advisory Services at EY.
Despite this promise, the use of AI in finance is too hot to handle for many finance chiefs. A challenge is the need to understand the interrelationship between different AI solutions, says Ankur Agrawal, a partner in McKinsey’s Corporate Finance and Health practice. “Gen AI alone is not the gamechanger in finance it’s made out to be. The power of technology to transform the finance function is a combination of automation, predictive AI and generative AI, not one or the other,” he explains.
Asked for an example of how the different technologies would combine to provide value in the finance function, Agrawal cited the use of AI in accounts receivable and payable. “Assuming automated AR and AP processes, generative AI can be used as an accelerant, segmenting the analytics produced by predictive AI to automatically send different email templates to different customers at different points in time,” he says.
A newer customer that’s behind on a payment by a few days would automatically receive a specifically tailored email alert, whereas a longstanding customer with highly favorable contract terms in the same situation might not receive notice of an overdue payment until months later. “Each month that goes by without payment would spur the generative AI tool to create and send a new set of emails to different customers,” Agrawal says.
“The beauty of combining all three technologies is repetition,” he says, explaining that the data produced in the successive machine interactions yields unique insights into customer payment behaviors. “By receiving early warning signals on customers with potential credit issues, you’re able to better predict the timing and amount of cash inflows, outflows and balances, improving cash flow forecasting,” he says.
Numbers Game
While there are advantages to be seized by CFOs in being a first mover with a new technology, the disadvantages of being a trailblazer racing forward to outdo one’s peers often outweigh them. “The CFOs I talk with are spending a lot of time thinking about AI to understand the use cases,” says Steve Gallucci, Global CFO Program Leader at Deloitte, adding that his conversations suggest that “most would be very happy to be a fast follower.”
Gallucci’s colleague Ranjit Roa, U.S. Finance & Performance Leader at Deloitte, agreed. “It’s not that CFOs are on the fence about AI investments, they just don’t want to be first on the field,” he says. “They need to see a broader ecosystem of AI uses in finance to figure out the risks. Right now, they have significant trust concerns over misleading outputs that are automatically generated and end up in reports. This doesn’t mean they’re not interested, however.”
Such interest has yet to culminate in a significant investment. “Every CFO I talk with is under a lot of pressure to modernize finance through technology, since almost every single finance process can benefit from AI technology in some fashion,” says EY’s Corson. “Reality is a bit different from the promise, however.”
While many CFOs have made investments in predictive analytics, the solution involves human interactions to input and query data, identify patterns and trends, and test the assumptions. Predictive AI, which is completely autonomous, takes the human out of the picture. That’s troubling for many finance chiefs.
“CFOs are concerned about the data quality of the outputs because the numbers have to be accurate all the time, not mostly accurate or accurate here and there,” says Corson. “Over the past year, we’ve seen generative AI outputs that couldn’t be tied back to a particular set of numbers or analyses.” To ensure accuracy, the financial outcomes generated by AI solutions need to be auditable, he says. “If and when this occurs, CFOs will feel greater comfort using the tools.”
Gallucci agreed. “The stewardship role of CFOs requires accurate reports. No CFO is ready to turn that over to a machine or a model that potentially generates misleading outcomes,” he says.
This trepidation weakens the cost-benefit analysis of the value of AI within finance. “It’s a fairly sizable investment. It’s not something where you can easily do a pilot project to figure out if it works,” says Roa. “Issues like trust, controls and accountability for hallucinatory outcomes must be resolved first.”
Agrawal offered a similar perspective. “The fact that you have `no throat to choke’ if the AI gets it wrong is a big problem,” he says. “It’s a combination of the fear of the unknown and undefined or no accountability. It’s not just the technology; it’s about people and processes. CFOs understand that more than other [business leaders].”
Hesitancy in implementing AI is not confined to finance chiefs alone. Corson pointed out that many best-of-breed software providers to the finance function have been slow to embed generative and predictive AI in their applications. “A lot of CFOs that want to deploy AI in finance have had to build their own solutions,” he says. “We’re beginning to see a few vendors embed AI in their products, the case with BlackLine’s intercompany AI solution to predict reconciliation issues and transaction failures, and with HighRadius’ AI-powered invoice processing solution.”
Other challenges include a severe shortage of skilled technical talent available to develop, use and ultimately vet AI solutions to ensure the outputs are accurate and auditable. “CFOs know they need to build these varied capabilities, but for now they’re seen as inhibitors,” Gallucci says.
Tentative Forays
The experts’ views largely coincide with the cautious approach taken by CFOs Robertson, Damiani and Ebeling. Damiani’s minimal investment in generative AI at Balentine has been “productive and helpful,” he says, “but it’s too early to say we’ve actually seen significant gains in time saved. We still have to double-check the outputs to ensure they’re correct. And we worry about data integrity. Our thinking is it will pay dividends in time.” The wealth management firm has 55 employees and $7.5 billion in assets under administration.
Asked about the potential use of predictive AI in the finance function, Damiani says he’s interested but his concerns outweigh his optimism. “As a CFO with a really good controller, predictive AI would create bandwidth for both of us to do more strategic things,” he says. “However, nothing has come across my radar that doesn’t raise red flags over data integrity and security. We’re fiduciaries; we can’t simply take it for granted it’s safe.”
Although CFO Ebeling at York Solutions is holding back on more robust investments in AI for the time being, she is nonetheless excited about the opportunities available. “We’re looking to align predictive AI and generative AI to do things like track our billing, produce financial reports, and leverage algorithms to make quicker and better forecasts,” she says. “If we can find a way to bundle all that together, it would be fantastic, getting me and the team out of the trenches studying the numbers to make more informed strategic decisions.” The IT consulting firm has 600 employees serving Fortune 500 companies primarily in the healthcare, financial services and manufacturing sectors.
At Rogers-O’Brien, CFO Robertson remains bullish on AI, despite his intention to avoid the bleeding edge. In addition to leveraging predictive analytics to identify and mitigate project completion issues, the finance organization uses Microsoft’s generative AI tool Copilot to search emails and MS Teams transcripts to develop content relevant to the finance function. The general contractor of commercial construction projects employs more than 500 people and tallied over $872 million in 2023 revenue.
“What we’re doing now is preparing us for the next steps,” he says. “We’re gathering data differently than we did in the past. We now require monthly project status reports that are more than just a bunch of numbers, using generative AI to interpret what the numbers mean in words. In the past, we had to do the interpretation.”
Robertson expects he will eventually use predictive AI to predict problems that cause a future delay in a project’s completion, such as a supply chain slowdown that results in only half the number of windows showing up at the construction site. “We know that if all the windows aren’t on site as scheduled, there’s an 80 percent chance of a project delay,” he says. “If the AI can suggest what we can do in advance of a supply chain slowdown to avoid this possibility, it would be a gamechanger.”
CFO Ann Anthony at Oberon Fuels, a producer of low carbon-intensity renewable fuels, has an advantage the other three finance chiefs lack. The late-stage startup is just putting together its finance tech stack, which will include both predictive and generative AI solutions. “From a back-office perspective, the function I lead is very lean, with just myself, the controller, a financial modeling guru and a part-time clerk,” Anthony says.
She is in the thick of plans to deploy an ERP system by the end of the year or early next year, just in time to address the company’s fast-track scaling. Partnerships and/or investments have been secured with Volvo Trucks, Mack Trucks and Ford, and more than 450 customers in the fuel and energy sectors have already lined up to buy its products.
“My plan is to go big with more technology than what we need at the moment,” says Anthony. “As I look around the corner at the ever-more strategic role of the CFO, I’m very focused on acquiring both predictive and generative AI forecasting models for use across the finance value chain.”
As other finance chiefs pull the lever on such actions, their peers inevitably will become fast followers. As Agrawal puts it, “The next AI versions will get better and better at transforming finance, turning those CFOs thinking about AI into confirmed users.”
Russ Banham is a Pulitzer-nominated business journalist and best-selling author.