Imagine pouring your AI budget into pilots and tools, yet your quarterly results show nothing. That's exactly what thousands of CEOs are experiencing right now.
What's Actually Happening?
In February 2026, the National Bureau of Economic Research published findings from a survey of 6,000 executives across the US, UK, Germany, and Australia. The verdict was striking. Nearly 90% of firms reported that AI has had no impact on employment or productivity over the past three years.
About two-thirds of respondents said they use AI, but actual usage amounted to just 1.5 hours per week. Meanwhile, 374 S&P 500 companies mentioned AI positively in earnings calls — yet the macroeconomic data shows no productivity bump.
Apollo chief economist Torsten Slok put it bluntly:
AI is everywhere except in the incoming macroeconomic data. You don't see AI in the employment data, productivity data, or inflation data.
Sound familiar? It should. Nobel laureate Robert Solow said almost the exact same thing about computers back in 1987.
Why This Is Different — And Why It Isn't
The Solow Paradox first appeared in 1987. IBM revenues had nearly tripled over the decade as companies raced to buy computers, yet productivity growth fell from 2.9% (1948–1973) to just 1.1% after 1973. Solow wrote in the New York Times: "You can see the computer age everywhere but in the productivity statistics."
The paradox had three root causes — and all three are playing out again with AI today.
| Root Cause | 1980s Computers | 2020s AI |
|---|---|---|
| Measurement gaps | Service quality gains invisible to GDP | Time saved by AI spent on more tasks (unmeasured) |
| Adoption lag | 30 years from circuits to productivity gains | AI pilots → process redesign takes years |
| Org. inertia | Existing workflows + computers bolted on | Existing workflows + AI bolted on |
Yet there are early signs of a reversal. Stanford economist Erik Brynjolfsson estimates U.S. productivity jumped 2.7% in 2025 — nearly double the prior decade's average. He sees this as the start of the J-Curve's upswing, a transition from AI investment to "harvesting" its benefits.
MIT Nobel laureate Daron Acemoglu is more cautious, projecting a modest 0.5% productivity gain over the next decade:
I don't think we should belittle 0.5% in 10 years. That's better than zero. But it's just disappointing relative to the promises people are making.
The Action Plan: How to Beat the Paradox
The lesson from the 1980s is clear: companies that simply bolted computers onto existing workflows gained little. Those that redesigned their operations around the technology — like retailers who used barcode data to overhaul their entire supply chains — captured all the gains. The same principle applies to AI.
- Set measurable outcomes before deploying AI
Not "feels faster" but "contract review time drops from 2 days to 4 hours." If you can't measure it, you can't manage the J-Curve. - Redesign the process, then add AI
Applying AI to a broken workflow just accelerates the broken workflow. Map your process first, cut unnecessary steps, then introduce AI as the enabler. - Limit active AI tools to three or fewer
BCG research found productivity peaks with 1–3 AI tools and drops sharply at 4+, due to cognitive overload they call "AI brain fry." - Budget for complementary investment
Brynjolfsson's research suggests firms need up to 10x their technology investment in training, org redesign, and governance to actually unlock productivity. - Manage expectations for the J-Curve dip
Early-stage performance plateaus are not failures — they're the structural dip before the upswing. Cutting investment at this point is the most common way organizations never complete the curve.
Go Deeper
NBER Working Paper #34836 The primary source: survey data from 6,000 C-suite executives across four countries on AI usage and outcomes. nber.org
The Productivity J-Curve: How Intangibles Complement General Purpose Technologies Brynjolfsson's foundational MIT paper explaining why productivity dips before it rises with transformative tech. ide.mit.edu
Back to the Future: Solow's Productivity Paradox in the Age of AI A clear-eyed comparison of 1980s IT adoption patterns and today's AI rollout, with practical organizational lessons. linkedin.com
The Productivity J-Curve and the Hidden Economics of AI Transformation Why 80–90% of transformation capital goes to technology but the real ROI driver is people and org redesign. medium.com
Erik Brynjolfsson on AI Productivity Liftoff (Fortune, Feb 2026) The Stanford economist's analysis of why 2025 may mark the inflection point in the AI productivity J-Curve. fortune.com
