CFOs believe AI is paying off. Researchers aren’t so sure—yet

CFOs Believe AI Is Paying Off. Researchers Aren’t So Sure—Yet

A recent study indicates that while executives recognize gains from AI, these improvements have not yet been reflected in financial performance.

Good morning. Artificial Intelligence (AI) is enhancing worker productivity; however, financial outcomes have yet to align with these advancements.

"Artificial Intelligence, Productivity, and the Workforce: Evidence from Corporate Executives" is a new working paper developed by researchers from Duke University’s Fuqua School of Business and the Federal Reserve Banks of Richmond and Atlanta. The study reveals that although CFOs report productivity increases attributed to AI, evidence based on revenue suggests a more measured impact—at least for the time being.

Drawing on surveys from nearly 750 executives, the research highlights a “productivity paradox.” Corporations reported average AI-driven productivity gains of approximately 1.8% in 2025, but when the researchers assessed actual revenue and employment data for implied gains, the findings demonstrated significantly lower increases across all major industries for both 2025 and 2026.

“It’s not really hitting the top line yet in full force,” remarked John Graham, a finance professor at Duke’s Fuqua School of Business and a co-author of the study. “There is some level of delay in here for sure.”

Timing Issues
Graham largely attributes this delay to timing. Companies that increased their AI investments in late 2025 have yet to fully implement capabilities, adjust pricing, or achieve revenue growth. The productivity gains reported for 2025 closely align with revenue-implied gains anticipated for 2026, suggesting a potential one-year lag in realization.

This phenomenon echoes the “productivity paradox” originally outlined by economist Robert Solow in 1987, which noted widespread computer usage that was not immediately visible in productivity data. The authors propose that AI may be undergoing a similar process.

Industry Disparities
The productivity gains from AI are not uniform across industries. High-skill services such as finance exhibit the most significant growth, while sectors like manufacturing, construction, and low-skill services show lesser, though still positive, advancements. These variations reflect the differing applications of AI depending on sector and company type.

“For some industries, AI will replace the call center,” Graham stated. “For others, it’s about optimizing conveyor belts in factories or reducing the number of financial analysts.”

Investment Justification
For CFOs, justifying AI expenditures before tangible returns materialize poses a challenge. “ROI often depends on exactly how you calculate it,” Graham noted. “What you’d really want is a multi-year perspective on value creation rather than merely a point in time.”

Graham underscores the necessity of demonstrating value over a three to four-year timeframe to avoid being ensnared in trends that may not yield clear benefits for the organization.

In summary, while AI holds the potential for substantial productivity gains, realizing these benefits in revenue is a complex journey that may require time. Factors such as implementation timelines and industry-specific differences play a crucial role in this process.

Big Deal: U.S. Workers' Well-Being Declines
A recent report from Gallup highlights a decrease in the well-being of U.S. workers since 2022, accompanied by a notable shift in perceptions regarding the job market. In Q4 2025, only 28% of U.S. employees believed it was a favorable time to secure a quality job, a drop from nearly 70% in mid-2022—marking the most significant decline in job market confidence recorded by Gallup in the past four years.