The AI productivity puzzle

When it comes to AI ROI, many CFOs are in a tight spot.

“Boards are putting pressure on CFOs to say more about or demonstrate more about the impact of AI adoption across the enterprise,” Randeep Rathindran, managing VP at Gartner, whose research focuses on CFOs and FP&A, told CFO Brew. A Gartner survey released in late October, in fact, shows that 77% of board members want to see cost savings and productivity gains from AI.

But at many organizations, those gains and savings have yet to occur, Rathindran said. A Gartner analysis of earnings calls, Rathindran said, showed that out of over 1,000 S&P 1200 companies, only 4% mentioned material cost savings due to AI in the first half of 2026.

Mature enough? What’s going on, and what can CFOs do about it? Part of the issue, Rathindran said, is that many companies are still in earlier stages of AI maturity. Companies that are further along in their AI deployments are “replacing legacy software” with AI and rolling out agents, he said, but most organizations are using AI for “more basic use cases” such as generating reports and automating transactional work. Many companies, he said, are still focused on getting their data AI-ready and training their AIs.

AI is thus creating productivity “at a task level” and not at the team level—and it’s hard to assess how much time is being saved on those individual tasks. The data’s “more qualitative than quantifiable,” Rathindran said.

And processes have yet to change to incorporate AI. “We’re either not changing the way we work as a whole, or we’re not putting [in] the right kind of operating models, or hiring for the right kind of roles,” he said.

Don’t expect headcount cuts. CFOs should also “temper their expectations” of headcount reductions due to AI, Rathindran said.

Another Gartner survey of 1,303 functional leaders from January through April 2026 found that 23% said their functions’ staffing levels had decreased after implementing AI, but 17% said their headcount had increased. (And 60%, flummoxed, replied that they “didn’t know” how much impact AI had on staffing.) AI can sometimes create more work, Rathindran said, by producing more outputs, such as analyses and forecast scenarios, which staff then have to review.

Ultimately, he said, AI should be viewed as a long-term investment. Some AI use cases, especially customer-facing ones, he said, are “long-lead-time activities.” Expecting a normal 12 to 24 month payback period for AI, as you would for a regular investment, is not realistic.

That can put CFOs in a bind, he admits: “It’s a very hard thing for CFOs because you can’t just sign a blank check and say, ‘Oh, well, five years from now things will be better.’”

The “portfolio” approach. In the meantime, there are things CFOs can do in the short run, Rathindran said. For example, they can work to gain greater visibility into AI use at their organizations, which may unearth opportunities for savings. A bank used AI to save time in its legal department, reducing the need for external legal review and contracting; it was able to insource that work and save on outsourcing, he said.

Having an enterprise-wide AI strategy with visible connections between AI use in different departments can help companies see more value from the technology. Take a “portfolio-level” view rather than “just experimenting in a lot of one-off use cases,” Rathindran advised. What really matters is whether your AI deployments reinforce each other. “Is what you’re doing in the product organization,” he said by way of example, “resulting in more customer insights that can be transferred to your R&D or research organization?”