SUBSCRIBE

A new NBER survey of 6,000 executives says 9 in 10 saw zero AI effect on their firm. The same paper's authors are forecasting real effects within three years.

The no-effect number is real and verified. It is also half of a two-part study, and the retelling online drops the other half.

01THE CLAIM
"A new NBER survey of nearly 6,000 senior executives (US, UK, Germany, Australia), fielded Nov 2025-Jan 2026, found more than 90% report no effect of AI use on their own firm's employment over the past three years, and 89% report no effect on labor productivity (measured as sales per employee)." [SOURCE ↗]

THE MOVE: CHERRY-PICKED SLICE, the flattering subset, presented as the whole

TRUE, BUT5 SOURCES · LIVE 2026-09-05
NBER WORKING PAPER 'FIRM DATA ON AI' TRACK RECORD1 CLAIM · 40/100 BS RATE →
~6,000senior executives (CEOs, CFOs, senior finance managers) surveyed across US, UK, Germany, Australia
~3,000additional US workers surveyed separately on employee-side expectations
Nov 2025-Jan 2026survey fielding window (paper revised March 2026)
>90%share of executives reporting no effect of AI use on own-firm employment over past 3 years
89%share of executives reporting no effect on labor productivity (self-reported sales per employee) over past 3 years
69%share of firms reporting current AI use (US 78%, Australia 59%)
1.5 hrs/weekaverage AI usage intensity among executives who use it regularly
+1.4%executives' forecast of AI's productivity boost over the NEXT 3 years
-0.7%executives' forecast of AI's employment reduction over the NEXT 3 years
+0.8%implied net output gain from executives' 3-year forward forecast
~1.75 millionjobs NBER estimates could be affected across the four countries by 2028, per forward-looking forecast
14%productivity gain (issues resolved/hour) found in a DIFFERENT NBER paper studying 5,179 customer-support agents using generative AI (34% for novice workers)
70%/80%alternate rounding of the same stats used in some secondary coverage (vs. 69%/89-90% in the primary paper) - shows popular retellings drift from the exact figures
02THE CHECK

THE CLAIM. NBER working paper w34836 surveyed nearly 6,000 senior executives across the US, UK, Germany, and Australia and found more than 90% report no effect from AI on their own firm's employment over the past three years, and 89% report no effect on labor productivity.

THE CHECK. both figures check out against the primary paper and match independent coverage from the NBER Digest and The Register. What the popular retelling in this record strips out is the other half of the same survey: the same executives forecast a real productivity gain and a real employment drop over the NEXT three years, and the productivity figure being measured backward is a single self-reported metric, sales per employee, not an audited output measure.

THE TWIST. a separate NBER paper on 5,179 real customer-support agents found a 14% productivity gain from generative AI tools, proof that a firm-level gut check can miss effects a task-level study catches directly.

03SAY THIS IN THE MEETING
"The '90% of executives saw no AI effect' stat is real. What gets left out is that the same executives, in the same survey, forecast it will hit within three years."

NBER working paper w34836, 'Firm Data on AI,' surveyed nearly 6,000 senior executives (CEOs, CFOs, senior finance managers) across the US, UK, Germany, and Australia, fielded November 2025 through January 2026 and revised in March 2026. Separately, about 3,000 US workers were surveyed on their own e

🔒 THE FULL AUTOPSY · FREE WITH AN ACCOUNT

You just read the free check. Sign in free, a code by email, no passwords, and the rest unlocks: the evidence trail, the steelman and the rebuttal, all 5 sources with quotes and screenshots, and our on-record call.

This story is a stable, citable object. If you can falsify a verdict, tell us. Corrections are loud here.