Companies around the world poured more than $250 billion into artificial intelligence in 2024. Yet for most of them, the returns remain almost nil, and AI productivity fails to meet expectations as they crash.
A major new study from the National Bureau of Economic Research surveyed nearly 6,000 CEOs, chief financial officers, and senior executives across the United States, the United Kingdom, Germany, and Australia. The findings landed like a cold splash of water on boardroom optimism. More than 80% of those leaders reported a negligible measurable impact on AI productivity or employment over the past three years. Some analyses of the data push that figure even higher, toward 90%.
A 40-year-old puzzle resurfaces
The gap between AI’s visibility and its economic footprint has led economists to look back into history for answers. Apollo chief economist Torsten Slok framed the disconnect with a direct reference to Nobel Prize-winning economist Robert Solow, who identified a strikingly similar pattern in the 1980s during the early computer boom.
“AI is everywhere except in the incoming macroeconomic data,” Slok wrote recently, updating Solow’s famous observation for the current moment.
Solow originally wrote, “You can see the computer age everywhere but in the productivity statistics.”
Those words, penned decades ago, now read like a template for 2026.
The parallel runs deep. Just as companies invested heavily in personal computers throughout the 1980s and saw little initial return, today’s enterprises spend aggressively on AI tools while efficiency curves stay flat. Productivity eventually surged years after widespread computer adoption. Whether artificial intelligence follows the same arc remains the central question dividing economists right now.
AI productivity dream fades as real data tells a different story
Part of the explanation lies in how shallowly most organizations actually deploy these tools. About two-thirds of executives reported using AI, but that usage amounted to only about 1.5 hours per week, and 25% of respondents reported not using AI in their operations at all.
That picture matches findings from the PwC 2026 Global CEO Survey, which covered 4,454 chief executives across 95 countries. Fifty-six percent said they had gotten “nothing out of” their AI investments. Only 12% reported that AI both grew revenues and reduced costs.
So while AI dominates earnings calls and marketing materials, the day-to-day reality inside most organizations tells a different story.
Task-level gains fail to reach the bottom line
The frustrating wrinkle in all of this is that AI clearly works when researchers study specific tasks in controlled settings. A Harvard Business School and BCG study found that consultants using GPT-4 completed 12.2% more tasks, 25.1% faster, at 40% higher quality. A Stanford and MIT study of customer service agents showed a 14% productivity increase, with the sharpest gains among less experienced workers.
Individual task improvements ranging from 14% to 55% appear consistently across studies. Yet those numbers never travel upward into organizational metrics or national economic data. Researchers now describe this gap as the central puzzle of the current AI moment.
Duke finance professor John Graham, a co-author of a related study, put it plainly: “It’s not really hitting the top line yet in full force. There is some level of delay in here for sure.”
Trust collapses even as usage climbs
Workers themselves add another layer of complexity. ManpowerGroup’s 2026 Global Talent Barometer, covering nearly 14,000 workers in 19 countries, found that regular AI use increased 13% in 2025, but confidence in the technology’s utility dropped 18%.
Employees keep using these tools. They just trust them less. That contradiction limits how deeply companies can embed artificial intelligence into core operations, particularly where accuracy and reliability matter most.
Too many tools create new problems
Research also suggests that piling on AI tools backfires. Workers using three or fewer applications reported stronger AI productivity outcomes. Those juggling four or more described mental overload, fragmented workflows, and a higher rate of small but costly errors. Experts now call this dynamic AI fatigue, a condition where the tools meant to simplify work instead multiply friction.
The case for patience
Not everyone reads the current data as a failure. Some economists point to emerging signals that suggest AI productivity gains may be building beneath the surface. Harvard economist Jason Furman calculated that AI-related investment accounted for 92% of U.S. GDP growth in the first half of 2025. Separately, economist Erik Brynjolfsson’s analysis found U.S. productivity grew roughly 2.7% in 2025, nearly double the 1.4% annual average of the past decade.
Brynjolfsson frames this through what he calls the J-curve hypothesis. General-purpose technologies tend to suppress measured productivity during the heavy investment phase. Gains then accelerate sharply once businesses finish rebuilding workflows around the new technology.
That pattern matches what happened with computers. The productivity surge that Solow never saw in the 1980s eventually arrived in the 1990s, once businesses figured out how to actually reorganize around the tools rather than simply add them on top of existing processes.
The real gap is organizational, not technological
Increasingly, experts agree that the AI productivity problem has little to do with the technology itself. Only 2% of European enterprises place AI at the CEO level. In most organizations, fewer than half of all roles have changed in any meaningful way to reflect AI adoption. Companies buy the tools. They rarely redesign the work.
The businesses seeing real returns treat AI as a structural change rather than a software subscription. Those that simply hand employees a chatbot and call it a transformation are the ones generating the headlines about disappointing results.
Artificial intelligence dominates the conversation in every industry. The data, for now, tells a more grounded story. Whether this decade ends with a productivity revolution or a very expensive lesson in hype depends entirely on what companies do next.
Do you see AI productivity gains in the near future, or are expectations permanently ahead of reality? Please share your thoughts below in the comments.