The NBER just asked ~6,000 CEOs across four countries the same question. 89% report zero productivity impact from AI over three years · and forecast +1.4% over the next three. Same people. Same survey. The gap is the story.
The NBER just surveyed 6,000 executives across four countries. 69% use AI. 89% say it did nothing. The same executives forecast a +1.4% gain next year · the same bet that just failed its first run. Here's what the data is actually saying · and the one framework that separates the 5.5% who got paid from the 94.5% who didn't.
69% adoption. 89% zero impact. 90% zero employment change. NBER w34836, Feb 2026.
The same executives now forecast +1.4% productivity, -0.7% employment. Why the bet runs again.
Theory of constraints. Why 5.5% got paid · and the other 94.5% never crossed the P&L.
Constraint · throughput metric · active-user denominator. Three questions before any AI renewal.
Same people. Same quarter. Same budget line. Opposite signal. The +1.4% forecast isn't a bet on better technology · it's a bet on better deployment.
FRAMEWORK · "REALIZED IMPACT" VS "FORECAST CONVICTION"Executives report little own-firm impact of AI over the last 3 years, with nine-in-ten reporting no impact on employment or productivity. These same executives predict sizable effects over the next 3 years · productivity up 1.4%, output up 0.8%, employment down 0.7%.· Yotzov, Barrero, Bloom, Bunn, Davis, Foster, Jalca, Meyer, Mizen, Navarrete, Smietanka, Thwaites, Wang · NBER WORKING PAPER W34836 · PUBLISHED FEB 2026 · REV. MAR 2026
4,454 CEOs. 56% say AI has delivered nothing. Only 10–12% saw any revenue or cost impact at all.
95% of generative-AI pilots failing. The failure mode isn't the model · it's the gap between the tool and the process trying to adopt it.
Executives forecast employment down 0.7%. Employees at the same firms forecast +0.5%. The workers have the better track record this cycle.
The escape path isn't more budget · the 2023–26 cohort already tested that. It's constraint alignment: deploy AI at the single step whose speed decides the quarter, not the periphery where 94.5% of the industry is still spending.