AI productivity paradox: executives report minimal real gains

Summary

New data from large executive surveys highlight a widening gap between the promise of artificial intelligence and its measurable impact on businesses. While adoption is high and frontier systems are achieving impressive scientific feats, most firms say AI has yet to move the needle on productivity, employment or earnings.

At the same time, some economists point to early signs that productivity is starting to improve, keeping alive the debate over when – and for whom – the AI revolution will show up in the numbers.

In practice

A recent National Bureau of Economic Research study of roughly 6,000 executives across countries including the United States, United Kingdom, Germany and Australia found that more than 80% of firms reported no impact from AI on either employment or productivity over the past three years. Most executives say they use AI, but their personal usage averages just 1.5 hours per week, and a sizable minority report not using AI tools at work at all.

A separate PwC Global CEO Survey of 4,454 leaders in 95 countries found that 56% of organizations saw neither increased revenue nor reduced costs from AI implementations over the past year. Only 12% of CEOs said their companies enjoyed both higher revenue and lower costs thanks to AI, suggesting a small vanguard is managing to turn experiments into real financial returns.

Despite this track record, expectations remain upbeat. In the same NBER research, firms forecast that AI will boost productivity by 1.4% and increase output by 0.8% over the next three years, even after largely flat results so far. Echoing an earlier era, Apollo Global’s chief economist summarized the dilemma by saying that AI is “everywhere except in the incoming macroeconomic data,” recalling Robert Solow’s famous line about computers.

At the cutting edge, AI’s raw capabilities continue to advance quickly. OpenAI says its GPT-5.2 model independently derived and formally proved a new formula in theoretical particle physics, solving a gluon scattering problem that had resisted researchers for more than a decade. Verified by teams at top universities, the result underlines how fast AI is progressing in specialist domains compared to its slower diffusion into everyday business workflows.

Context

The tension between rapid technical progress and sluggish economic results has revived talk of a modern productivity paradox for AI. Much like the early decades of the personal computer, many experts argue that organizations need time to re-engineer processes, train people and redesign work before headline productivity statistics catch up with technological change.

Workplace research points to another wrinkle: AI tools may be intensifying work rather than lightening it. An eight‑month study from UC Berkeley’s Haas School of Business found that AI “consistently intensified work,” expanding employees’ task lists and eroding natural pauses during the day. That may help explain why workers feel busier and more assisted by AI without seeing clear gains in pay or performance metrics.

Still, optimists such as Stanford economist Erik Brynjolfsson argue that a recent pickup in US productivity growth could signal the start of a “harvest phase” for AI investments. For now, though, the central question remains unresolved: when will AI’s dazzling technical progress finally show up in the broad economic indicators that executives and policymakers care about most?

Why it matters

• Shows that simply rolling out AI tools at work does not automatically translate into higher productivity, revenue or cost savings.

• Helps employees and managers calibrate expectations about what near‑term AI projects can realistically deliver.

• Highlights a growing divide between a small group of AI “vanguard” firms and the many companies still struggling to see concrete returns.

• Suggests that realizing AI’s promise may depend as much on organizational change and job design as on smarter models.