AI Productivity Paradox: 90% of Firms Report No Impact, Echoing 1987 Solow Warning
Summary
- • • NBER survey of 6,000 executives finds nearly 90% of firms report AI had no impact on employment or productivity over three years
- • • Executives use AI only ~1.5 hours per week; 25% report not using AI at work at all
- • • Apollo chief economist: 'AI is everywhere except in the incoming macroeconomic data'
- • • Economists are invoking the Solow Productivity Paradox — the same pattern seen when computers failed to lift 1970s productivity
Details
NBER: 90% of firms report zero AI impact on employment or productivity
A February 2026 NBER study of 6,000 executives across the US, UK, Germany, and Australia found nearly 90% of firms saw no AI impact on employment or productivity over the prior three years. About two-thirds reported using AI, but averaged only 1.5 hours per week; 25% said they do not use AI at work at all.
Economists invoke Solow Productivity Paradox — history appears to be repeating
Robert Solow's 1987 observation that computers were 'everywhere except in the productivity statistics' is being directly applied to AI today. Productivity growth slowed from 2.9% (1948–1973) to 1.1% after the computer revolution — economists see an analogous pattern emerging with AI.
Apollo chief economist: AI absent from all major macroeconomic indicators
Apollo chief economist Torsten Slok wrote that AI does not appear in employment, productivity, or inflation data. He noted that outside the Magnificent Seven, there are no signs of AI in profit margins or earnings expectations despite record investment levels.
$250B+ in corporate AI investment in 2024 with no matching macro productivity gains
Corporate AI spending exceeded $250 billion in 2024, with US business AI spending projected at $200 billion for 2026. Despite this, no broad macroeconomic productivity signal has emerged, widening the gap between investment and measurable returns.
374 S&P 500 companies cited AI positively in earnings calls; gains not reflected in productivity data
A Financial Times analysis covering September 2024 to 2025 found 374 S&P 500 companies citing AI positively on earnings calls, yet this corporate optimism has not translated into measurable productivity improvements at the macro level.
Executive AI productivity forecasts remain optimistic despite current flat results
Despite reporting no current impact, surveyed executives forecast AI will increase productivity by 1.4% and output by 0.8% over the next three years. Executives expected a 0.7% employment cut, while individual employees projected a 0.5% employment increase.
St. Louis Fed: 1.9% cumulative productivity growth since ChatGPT; MIT projects only 0.5% over a decade
The Federal Reserve Bank of St. Louis observed a 1.9% excess cumulative productivity gain since late 2022. A 2024 MIT study offered a more conservative estimate of 0.5% over ten years. Nobel laureate Daron Acemoglu called even that modest figure 'disappointing relative to the promises.'
85% of professionals use AI but only for ~4% of actual work time
Hubstaff and Worklytics data shows high nominal AI adoption among professionals, but usage is concentrated in a narrow slice of tasks, leaving 96% of work entirely human. This shallow engagement helps explain why macro productivity gains have not materialized.
HBR (Feb 2026): AI raises work intensity rather than reducing workload, contributing to developer burnout
A Harvard Business Review study found AI tools are adding to workers' cognitive load rather than offloading it. Software developers were specifically cited as experiencing burnout, complicating the narrative that AI is a net benefit for knowledge workers.
Remote workers remain the clearest identified beneficiaries of AI productivity gains
Among workforce segments analyzed, remote workers show the largest measurable AI productivity improvements, suggesting that the benefits of AI tools are unevenly distributed across work contexts rather than broadly shared.
Research = study findings, Stat = quantitative data point, Context = direct attributed statement or framing, Insight = analytical argument, Industry Update = business/workforce trend
What This Means
The AI productivity story is now being explicitly framed by economists as a replay of the Solow Productivity Paradox — the same dynamic that saw decades of computer investment fail to lift macro statistics. With nearly 90% of firms in a major NBER executive survey reporting zero measurable impact, and Apollo's chief economist confirming AI is absent from every key macroeconomic indicator, the gap between corporate enthusiasm and hard data has become impossible to ignore. For businesses committing hundreds of billions to AI deployment, the strategic implication is stark: without a fundamental shift in how deeply and broadly AI is integrated into actual workflows — not just episodic 1.5-hour-per-week usage — the productivity payoff may remain elusive for years, much as it did with the personal computer.