Most AI initiatives produce unclear returns, finance execs say
Even as the use of AI within the finance function continues to normalize, the financial impact from most of the investments already made could be either positive or negative.
That’s because a majority of finance executives don’t know what return they’re getting on their AI spend for a majority of specific use cases, according to new research from Gartner.
In a survey of 160 senior finance-function leaders, 55% reported positive overall returns from their 2025 AI initiatives. However, “the picture becomes murkier at the individual use-case level,” Gartner wrote in its survey report.
That is, when asked about the ROI from the most-common AI use cases for finance, 57% of the use cases evaluated had unclear returns, while 18% were positive, 19% negative and 6% break-even.
Even among the 43% minority of cases for which there was clarity on return, the picture was mixed: 42% positive, 45% negative and 13% break-even.
“The contradiction between overall portfolio performance and individual use-case returns suggests that AI portfolios rely on asymmetric returns,” Gartner observed. “While a higher volume of AI initiative may fail to provide returns exceeding their costs, the successes of a few select initiatives are substantial enough to outweigh those losses.”
The situation puts CFOs in a difficult position, Gartner noted, as they must continue to allocate resources to AI pilots while acknowledging that a significant portion of that spend will be a sunk cost.
Therefore, CFOs should prioritize AI investments “based on strategic alignment, organizational readiness, and scalability” rather than solely on predicted ROI, according to Gartner. Further, they should learn when to cut their losses on underperforming initiatives through the use of performance targets and timelines for project termination.
However, different use cases can be expected to have different timelines. At the short end, AI projects aimed at improving numerical data extraction/retrieval on average took about 9.3 months to deliver the expected value, based on survey participants’ responses. At the long end were projects around scenario planning and simulations, which required 13.7 months on average.
About a third (35.3%) of finance functions’ AI spend was targeted at productivity improvements, such as business optimizations and employee hours spent per task. A significant share (20.9%) was devoted to cost reduction, such as supplier spend and headcount reduction.
Further, between 5% and 9% went toward each of several other goals: evolving the enterprise (e.g., new products or processes); managing risks, such as regulatory compliance); fostering resilience (operations, business continuity); increasing revenue, such as through market expansion; transforming the enterprise (e.g., a new business model); and retaining revenue, such as by improving customer service.
“AI adoption has reached the point where CFOs must adopt more deliberate portfolio management,” Gartner wrote. “The goal is not to stifle experimentation, but to know where to invest, when to cut underperforming initiatives, and which foundational capabilities to accelerate.”
