Firm data on AI
S · Shapes the field Nicholas Bloom, Steven J. Davis and central-bank survey teams (Atlanta Fed, Bank of England, Bundesbank) ask nearly 6,000 senior executives identical questions across four countries, a rare representative firm-level dataset on AI.
Summary based on the full text·Original: about 75 minutes to read
In short
NBER Working Paper 34836 (February 2026, revised March 2026) by Ivan Yotzov, Jose Maria Barrero, Nicholas Bloom, Steven J. Davis and colleagues at the Bank of England, Federal Reserve Bank of Atlanta, Deutsche Bundesbank and Macquarie University. The team asked nearly 6,000 senior executives in four countries how AI has affected their own firms and what they expect next. The headline: AI use is widespread, but nine in ten executives see no effect on employment or productivity so far. They expect modest gains in the next three years, with productivity up and headcount slightly down.
Key numbers
- 69%firms across the US, UK, Germany and Australia that currently use AI
- 1.5 hoursaverage weekly AI use by the executives themselves
- 89%firms reporting no AI impact on labour productivity over the past three years
- -0.7%employment change executives expect from AI over the next three years (employees expect +0.5%)
As reported by the source.
English·PDF·opens nber.org
What it reports
- 69% of firms use AI on average: US 78%, UK 71%, Germany 65%, Australia 59%.
- The most common use is text generation with LLMs (41% of firms); image processing with machine learning 20%, robotics 9%, autonomous vehicles 3%.
- In the UK, firms using at least one AI technology rose from 61% (February-April 2025) to 71% (November 2025-January 2026). 75% of firms expect to use AI within three years.
- 28% of executives do not use AI at all in a working week; 41% use it up to 1 hour, 24% 1-5 hours, 7% more than 5 hours. Average: about 1.5 hours a week.
- More than 90% report no AI impact on own-firm employment over the past three years; 89% report no impact on productivity (sales per employee). Implied past productivity effect: about +0.29%.
- Next three years: expected productivity +1.4% (US +2.3%, UK +1.9%, Germany and Australia +0.9%), output +0.8%, employment -0.7% (UK -1.4%, US -1.2%). The authors translate -0.7% into about 1.75 million fewer jobs by 2028 across the four countries.
- In a UK sub-sample, about two-thirds of the expected headcount reduction comes from hiring fewer new employees rather than exits.
- About 3,000 US employees asked the same questions expect AI to raise employment at their firm by about 0.5% and productivity by about 0.9%.
How they got it
Parallel survey questions were fielded between November 2025 and January 2026 through existing central-bank business panels: the Atlanta Fed survey in the US, the Decision Maker Panel in the UK, the Bundesbank panel in Germany and a Macquarie University survey in Australia. Respondents are mostly CEOs, CFOs and senior finance managers. Categorical answers are converted to numbers (±7.5% for large effects, ±2.5% for small, 0 for none). Employee views come from the US Survey of Working Arrangements and Attitudes.
Limits
- Effects are self-reported perceptions and forecasts, not measured outcomes; the conversion of categories to percentages is a modelling choice.
- The sample covers existing firms only, so jobs created by new firms and economy-wide effects of higher productivity are outside scope.
- At least one co-author disclosed additional relationships of potential relevance.
Why it matters for companies
Widespread licences and pilots have not yet produced measurable productivity at most firms, so a manager should expect a gap between adoption and results. Executives plan to reduce headcount mainly through slower hiring, which hits entry-level pipelines first. The gap between what leaders and employees expect on jobs is a communication risk that HR and line managers will have to handle. The country spread suggests US competitors will move faster than European peers.
Summary based on the full text. Read the original for the full argument and data.Open PDF on nber.org
- Tags
- AI adoption, Productivity