More than 90% of firms report no measurable headcount change from artificial intelligence over the past three years. The finding tempers fears of rapid labor displacement while signaling that enterprise AI remains in an early deployment phase.
The data comes from a National Bureau of Economic Research working paper produced by researchers at the Federal Reserve Bank of Atlanta, the Bank of England, the Deutsche Bundesbank, and Macquarie University. The study surveyed nearly 6,000 phone-verified executives across four countries, primarily the UK and Germany, and cross-checked responses against a decade of national output and employment data.
Will AI Productivity Accelerate Over The Next Three Years?
Adoption is already widespread. Around 69% of firms report using some form of AI, led by large language model-based text generation at 41%, machine learning for data processing at 28%, and visual content tools at 29%. In the UK, adoption rose from 61% to 71% across 2025. By contrast, the US Census Business Trends and Outlook Survey estimated AI use at 18% by December 2025, reflecting differences in sampling and definitions.
Executives expect stronger gains ahead. On average, respondents project a 1.4% rise in productivity and a 0.8% increase in output over the next three years, with US executives forecasting a 2.25% productivity gain. Headcount is expected to decline by 0.7% overall, largely through slower hiring rather than layoffs. In economies with a decade of weak productivity growth, even incremental gains compound meaningfully.
The study also highlights a perception gap. US employees surveyed separately expect employment at their firms to rise by 0.5%, while executives anticipate a 1.2% decline. Employees forecast smaller productivity gains than management. Does this divergence reflect different vantage points on how AI augments daily tasks versus reshapes cost structures?
Researchers note that general-purpose technologies often deliver gradual improvements before broader economic impact becomes visible. As deployments mature and integration deepens, the next catalyst will be whether projected productivity gains begin to register in national statistics rather than executive forecasts.
