Editor’s note: This is the second in a two-part report exploring the potential risks and benefits for CFOs as they incorporate artificial intelligence into company operations.
CFOs and engineers tend to use superlatives when describing the future impact of artificial intelligence, labeling it a more disruptive technology than the steam engine, telegraph, or mainframe computer, to name a few.
“It’s as big, if not bigger, than the dot-com Internet boom,” Bart Chao, chief financial officer at process automation provider Nintex, said in an interview, calling the pace of change “uncertain.”
Global spending on AI infrastructure is This year, it will increase by 53% to $487 billion.according to International Data Corporation. Spending is likely to grow at an average annual rate of about 31% over five years and exceed $1 trillion by 2029, according to IDC.
“This is not the time for analytics paralysis,” said Chad Gold, chief financial officer at behavioral data company Full Story. “Things are moving too fast.”
“That doesn’t mean you should open your checkbook and pour money into everything,” he said in an interview. “But now more than ever we need to be willing to let our teams experiment.”
Technologists and executives say the challenges of implementing AI will test CFOs’ flexibility and judgment more than any other traditional technology.
Financial and competitive risks are high, so CFOs need to carefully assess the risks of AI, financial executives and technologists said, pointing to seven dangers:
1. Low or no return on investment
For many companies, revenues from AI are lagging behind investments in the technology.
More than half of CEOs (56%) said yes Failure to generate revenue or achieve cost savings According to the results of the past 12 months of data from AI PwC survey 4,454 CEOs from 95 countries announced in January. (Only 30% reported an increase in revenue and 26% reported a decrease in costs.)
However, the potential benefits are clear among AI adoption leaders, according to McKinsey.
20 such things The company increased EBITDA by 20%McKinsey said, reaching breakeven within 24 months and generating $3 in incremental EBITA for every $1 invested.
According to McKinsey, AI leaders focused on no more than three of their companies’ business areas, maintained a “maniacal” focus on customers and AI users, and insisted on accountability to KPIs.
Most CFOs (71%) believe in a common thing ROI metrics are inappropriate A study published last month by EY found that there is enough to measure the benefits from AI and other emerging technologies.
Traditional ROI frameworks do not accurately measure future indirect or intangible benefits, such as improved decision-making, predictive accuracy, and operational agility, EY said.
According to EY, CFOs will benefit from qualitative measures such as AI improving pricing, streamlining supply chains, and freeing up finance employees to focus on higher-value tasks.
Gold said CFOs could also benefit from being patient.
“What you learn with AI is you have to be willing to invest time on the front end,” he said.
“Initial training of an AI can take more time than performing a task,” Gold said. “But if you invest your time in the right way, the returns can be quite significant in the long run.”
Technologists and finance executives said CFOs wary of attacks on AI can start with pilot projects or forge partnerships to reduce risk.
“You can share the ROI risk by entering into a contract with a service provider and setting the amount they are paid based on the profits they receive,” says Christopher Wright. Global CFO Solutions and Performance Improvement Leader at Protiviti; said in an interview.
2. Loss of organizational knowledge
Relying on AI to collect and analyze data risks undermining the knowledge employees have gained through years of problem-solving on core business subjects like financial planning, customer relationships, and risk management, technologists and CFOs say.
“If you ask AI to do that, it’s dangerous because all of a sudden your human capital is gone,” Gold said, noting that people who grew up using GPS often don’t know how to use paper maps.
CFOs may automate forecasts, but how do they replace the knowledge their teams gain by creating those forecasts: what they learned about the business and how they made changes along the way? Gold said.
Wright said the potential loss of institutional knowledge and sound judgment is one of several reasons to ensure strict human oversight of AI.
“You want to do it with some degree of autonomy, but you always need human judgment and double-checking,” Wright said. “A prime example of this is that you don’t want your agent to make the final payment decisions, but you also don’t want them to manually enter invoices.”
3. Little or no governance
Many executives are concerned that internal AI agents could run amok, spewing proprietary data far and wide, opening up vulnerabilities to cyberattacks, or promising customers benefits or services that don’t exist.
But engineers and CFOs say the biggest threat comes from humans.
“The biggest risk is not a fraudulent model,” says ModelOp CEO Dave Trier. “This is a cottage industry that most companies operate today, with each team building and deploying AI differently and with no common standards,” he said in an email.
Kfir Lipman, CFO of Salt Security, an AI security platform, also believes that the “shadow agent problem” is the biggest danger posed by AI.
“Teams across the organization are deploying AI agents on their own and connecting them to internal systems, customer data, and financial workflows without central oversight or budget approval,” he said in an email. “These are real operational and financial commitments that are made outside of normal governance processes.”
Four out of five organizations (82%) discovered that an AI agent was being used within their organization in the past year. they didn’t know existedsaid Lipman, citing research by the Cloud Security Alliance. “Each of these agents is a risk for non-budgeted items to the business and can have compliance liability.”
Sound governance of AI starts with CFOs asking basic questions, Wright said. Are systems such as enterprise resource planning software compatible with AI? Can this technology interact with corporate data? Do your employees have the skills and tools to work with agents? This technology increases your exposure to cyber-attacks. Is your company’s board ready to interact with AI?
Pipedrive, a provider of customer relationship management software, strengthened its governance by reviewing AI procurement proposals with teams in departments that oversee security, compliance, technology and privacy, said CFO Regi Vengalil. The team meets twice weekly to evaluate suggestions, primarily from product engineering and marketing staff.
“Over the past few weeks, we’ve been working on how we can make it faster while still maintaining the right constraints,” he said in an interview.
4. Defective data
Without accurate, timely and easily accessible data, AI adoption is likely to be slowed or halted, technologists and financial executives said.
“Some people are hoping that trash will turn into gold,” Chao said. “In time, it may become fool’s gold.”
Technologists and financial executives said cleaning up the data could be costly.
“The core question for us is how do we migrate much of our old code and architecture to a more agile and flexible architecture for future destinations,” Bengalil said.
“This is our biggest opportunity,” he said, flagging the challenge of finding weaknesses in data without spending excessive time and money.
Engineers and CFOs said it’s critical to ensure humans monitor data, noting that while AI is praised for quickly resolving problems, it can also quickly create problems.
“Having a process that actually enables observability is very important,” Chao said. “If you don’t actually monitor every step in between, or you don’t have the ability to inspect it, there’s a real risk of error propagation getting out of hand.”
Wright said AI is still in its infancy, and much of the technology suffers from an “educational gap” and needs human help.
“Until we train an AI model, the experienced judgment of people with organizational memory will probably be better than the AI model,” he says. “The machine has not yet learned what it is supposed to learn.”
5. “Black box” is unreliable
AI insights derived from clean data can be unconvincing and frustrating to executives if the underlying inferences aren’t clear, technologists and CFOs say.
“Imagine a user receiving recommendations from us. If they don’t understand where our opinions come from, they won’t trust it,” Bengalil said.
“It can’t be a black box and needs to provide reasonable traceability to users so they can see the data that supports the rationale for their recommendations,” he said.
Unexplained AI discoveries are often mixed with at least some opaque insights that need to be unraveled, Chao says. “Big problems are often a combination of smaller problems.”
Wright said finance departments should be wary of relying on AI for forecasting, a high-stakes production category that is susceptible to black-box issues.
“Financial institutions aren’t going to use AI to make predictions. AI doesn’t have the judgment that they do,” Wright said.
He said AI can help speed up information collection, format it for analysis, and provide a first draft of predictive insights. “But ultimately you need to fact-check and make sure you’re dealing with real data and not imaginary data.”
6. Compliance with numerous regulations
The CFO and a wide range of financial executives were selected. Data security and privacy are the biggest risks Protiviti’s global research finds that AI adoption is on the rise.
Companies that adopt AI enter the thicket of AI regulation.
Protecting yourself from such dangers is not easy. CFOs and technologists say governments have enacted overlapping and contradictory AI regulations, complicating compliance with standards around data privacy, cybersecurity, intellectual property and other key vulnerabilities.
To streamline AI implementation, CFOs and their C-suite colleagues need to include lawyers, compliance and cybersecurity experts in decision-making, Wright said.
Wright said in an interview that experts can help identify who within the company is responsible for ensuring compliance, as well as plan for training employees and hiring outside experts.
Pipedrive CFO Vengalil said compliance, legal and security experts “need to be built into the product build.”
Pipedrive, a provider of customer relationship management software with operations in multiple countries, must ensure its customer call records comply with privacy regulations, Chief Financial Officer Bengalil said.
“We are considering all the complex dynamics that arise as part of technology acceleration,” he said. “You don’t want to get too far ahead and find out at launch time that you can’t launch.”
7. Employee fear and resentment
Technologists, chief financial officers (CFOs) and others say employees often react to the introduction of AI in one of two ways: either resist it as a threat to their jobs or welcome it as an opportunity to improve their skills and keep pace with innovation.
“We risk losing people who don’t want to innovate or who are afraid of this development,” Chao said.
Elizabeth Ngonzi, an AI consultant and board member of the American AI Association, said companies that promote adoption often start from the top down without getting sufficient input from employees in finance, operations, human resources, and other roles.
“You end up automating broken workflows, hardcoding outdated assumptions, and deploying tools that no one trusts or uses,” she said in an email.
“Financial risk manifests itself in sunk implementation costs, slow adoption, and ‘AI on a slide deck’ rather than day-to-day decisions,” she said.
The solution is openness and collaboration with employees at all levels, engineers and CFOs said.
“You have to bring people along, listen to what people have to say, and see what hopes and fears people have about this and how that aligns with the path you’ve set out,” Chao said.
