Key Findings

  • A working paper released this month by the National Bureau of Economic Research (NBER) shows that surveyed CFOs believe AI did not have a substantial impact on employment at their organizations last year, and they expect AI's impact on total headcount to remain limited this year—but they anticipate an overall shift in the skill composition of the workforce.
  • Finance executives expect thatintroducing AIwill lead to a decline in "routine clerical positions" over the next three years, with the share of such jobs projected to fall by 2 percentage points by 2028. Meanwhile, the share of technical roles, including engineers, data scientists, and analysts, is expected to grow—rising by 0.62% in 2026 and 1.35% by 2028. The survey was conducted among 750 CFOs by economists at the Federal Reserve Banks of Atlanta and Richmond.
  • Notably, the paper also finds that, among existing firms, cost reduction—including both labor and non-labor expenses—is one of the least important motivations for AI investment. "Overall, the goal of investing in AI in the short term is not to cut headcount or costs, but to enhance productivity," the study notes.

Deeper Analysis

The study examines CFO responses on how AI use is helping to improve employee productivity, asking finance executives to report percentage changes over the past year and the next two years.

The survey shows that, in 2025, average labor productivity growth attributable to AI was 1.8%. Productivity gains are expected to "strengthen significantly" this year, with average reported labor productivity potentially reaching 3%.

"Labor productivity gains are positive and vary across industries, and are expected to strengthen further in 2026, with the largest impacts concentrated in high-skill services and finance," the study says. The survey covers four industries, including finance, manufacturing and construction, high-skill services, and low-skill services.

The study also asked CFOs to report percentage changes in revenue and employment due to their companies' use of AI, finding that these "implied productivity gains" are also positive, with implied labor productivity growth in finance at approximately 0.8%.

These findings come as employers and CFOs are placing greater emphasis on AI skills, particularly in finance: for example, a recent study by Datarails found that about one-third of finance job postingsnow explicitly mention AI skills, compared to just one-quarter a year ago, as CFO Dive has reported. Finding effective ways to implement AI within finance teams is also becoming a top priority for many CFOs: Hewlett Packard Enterprise CFO Marie Myers told CFO Dive last month that her finance team is seekingto make greater use of agentic AI tools。

The survey also shows that finance executives hold positive views on AI's future ability to enhance employee productivity, even though short-term gains may not yet be visible. The study identifies a so-called AI "productivity paradox," where respondents consistently report AI productivity gains that are "higher than those implied by contemporaneous changes in revenue and employment."

However, the study also notes that this gap may stem from "delays in output realization and quality improvements that have not yet been captured in measured revenue."

"More broadly, firms' conceptual understanding of productivity appears to extend beyond mechanical per-capita revenue calculations to encompass improvements in workflows, task efficiency, and organizational capabilities, the revenue effects of which will only materialize gradually," the paper states.