90% of CEOs Say AI Has Had Zero Impact on Productivity - Economists See a Familiar Ghost

Nearly 90% of firms say AI has had no impact on employment or productivity over the past three years. That is not a fringe opinion from a handful of skeptics. It is the headline finding from a major new study published by the National Bureau of Economic Research this month, surveying nearly 6,000 CEOs, CFOs, and senior executives across the United States, United Kingdom, Germany, and Australia.

The result lands like a cold shower on an industry that has been running hot on promises. More than $250 billion was poured into AI during 2024 alone. Three quarters of S&P 500 companies now mention AI on earnings calls. And yet, when researchers asked the people actually running these companies whether they could see it in the numbers, the overwhelming answer was: not yet.

The Solow Paradox Returns

Economists are reaching for a familiar metaphor. In 1987, Nobel laureate Robert Solow looked at the explosion of personal computers flooding American offices and made an observation that became one of the most quoted lines in economics: "You can see the computer age everywhere but in the productivity statistics."

What followed was nearly a decade of what researchers call the productivity paradox - massive IT investment, negligible measurable gains. Productivity growth had dropped from 2.9% between 1948 and 1973 to just 1.1% after 1973, even as computers became ubiquitous.

Now, nearly 40 years later, Apollo chief economist Torsten Slok sees the same dynamic playing out. "AI is everywhere except in the incoming macroeconomic data," he recently noted.

What the Data Actually Shows

The NBER paper, authored by a 13-person team including economists from Stanford, MIT, the Bank of England, and the Atlanta Fed, represents the first large-scale international dataset on firm-level AI use. Their findings paint a nuanced picture.

About 70% of surveyed firms actively use AI, with younger and more productive companies leading adoption. More than two thirds of executives report using AI personally. But dig into the how and the picture shifts: average executive AI usage amounts to just 1.5 hours per week. A full quarter of respondents say they do not use AI in the workplace at all.

Over 80% of firms report negligible effects on either employment or productivity over the past three years. The technology is present. The transformation is not.

The Optimism Gap

Here is where it gets interesting. Despite the current lack of measurable impact, these same executives remain bullish on the future. Firms forecast that AI will boost productivity by 1.4%, increase output by 0.8%, and cut employment by 0.7% over the next three years.

There is a catch, though. MIT economist Daron Acemoglu calls the projected 0.5% overall productivity increase "disappointing relative to the promises." And a notable gap exists between how executives and employees see the future: while management predicts job cuts, surveyed employees actually predict 0.5% employment growth - a meaningful disconnect between the boardroom and the people on the floor.

A Pattern Across Studies

The NBER paper does not exist in isolation. PwC's 29th Global CEO Survey, released at Davos in January, surveyed 4,454 CEOs across 95 countries and found that 56% of companies are getting "nothing out of" AI adoption. Only 10-12% report seeing benefits on either revenue or cost. PwC Global Chairman Mohamed Kande pointed to a root cause that has nothing to do with model capabilities.

"Somehow AI moves so fast that people forgot that the adoption of technology, you have to go to the basics," Kande said. Clean data. Solid business processes. Governance structures. The fundamentals that most organizations still lack.

A separate Workday study found that while 85% of respondents said AI saved them 1-7 hours per week, about 37% of that time savings was lost to rework - correcting errors, rewriting AI-generated content, and verifying output. And a METR study found that experienced open-source developers using AI coding tools actually completed tasks 19% slower than when working without them.

Why the Delay?

The historical answer offers both comfort and caution. The original Solow paradox eventually resolved itself. By the late 1990s, IT investments started paying off spectacularly as organizations learned to reorganize their workflows, business processes, and management structures around the technology. The payoff came - it just took about 15 years.

The same pattern may be unfolding with AI. ManpowerGroup's 2026 Global Talent Barometer found that across nearly 14,000 workers in 19 countries, regular AI use increased by 13% in 2025, but confidence in the technology's utility plummeted 18%. Workers are using AI more and trusting it less. That is not necessarily a bad sign - it might indicate a maturation from hype to realistic assessment.

But the scale of investment puts real pressure on the timeline. IBM's CHRO recently announced the company is tripling Gen Z entry-level hires, citing automation concerns. Some economists point to Q4 GDP tracking at 3.7% as a potential signal that productivity effects are starting to materialize. Others say we are still years away.

The Bottom Line

The NBER study does not argue that AI is overhyped or that it will never deliver. What it documents, with unusual rigor and international scope, is the gap between what we expect AI to do and what it has actually done so far. History suggests that gap will eventually close. But history also suggests it will take longer and look different than most boardroom presentations currently assume.

For the companies betting billions on AI transformation, the message from their own C-suites is clear: the revolution has been widely purchased but not yet installed.

Sources: