A firm survey: what AI has changed so far
In February 2026 NBER issued "Firm Data on AI": a survey of almost 6000 executives at firms in the US, UK, Germany and Australia. Around 70% of firms use AI and over 80% report no effect on employment or productivity over three years.
Why it matters
Instead of forecasts there is a measurement: how many firms in four countries use AI and what their own executives have seen over three years. An editorial assessment: it is executives' self-report, not a count of output.
What the February version says. Respondents are CFOs, CEOs and other executives from stratified samples; authors at the Bank of England, the Federal Reserve Bank of Atlanta, the Bundesbank and Stanford. Around 70% of firms actively use AI, more often younger and more productive ones; over two thirds of executives use AI regularly, on average 1.5 hours a week, a quarter never; over 80% of firms report no effect on employment or productivity over three years. For the next three years executives forecast +1.4% productivity, +0.8% output and -0.7% employment; surveyed employees expect +0.5% employment. What the record does not claim. The March revision changed three figures (69%, "nearly 6,000", nine in ten with no effect), so the above is the February version from the Internet Archive capture, not the page as it stands. Whether "nine in ten" is a different definition from "over 80%" the passages read do not say. Executives' forecasts are expectations, not a measured effect. NBER working papers are not peer-reviewed.