The AI Productivity Puzzle: What Corporate Executives Are Saying

AI adoption and its effects on firms: evidence from the US, UK, Germany, and Australia

How has AI affected productivity, employment, and output across major economies? This is the question that Yotzov et al. are answering in this paper.

The authors survey 6,000 senior business executives (CEOs, CFOs, and senior managers) at firms across the US, UK, Germany, and Australia, using four existing central-bank-affiliated panel surveys:

1. US: Survey of Business Uncertainty (SBU), fielded by the Federal Reserve Bank of Atlanta (November 2025)
2. UK:  Decision Maker Panel (DMP), run by the Bank of England (November 2025–January 2026)
3. Germany: Bundesbank Online Panel of Firms (BOP-F) (January 2026)
4. Australia: Business Outlook Scenarios Survey (BOSS), run by Macquarie University (December 2025)

The surveys asked executives about: (1) which AI technologies their firm currently uses, (2) how frequently they personally use AI, (3) the realized impact of AI on employment and productivity over the past three years, and (4) their expected impact over the next three years. An additional employee survey (SWAA) was used in the US to capture workers’ views separately and compare them to executives’. The surveys are validated against national GDP and employment aggregates, showing close alignment over a decade.

Findings

• Wide but uneven adoption: 69% of businesses across the four countries actively use AI, led by the US (78%), UK (71%), Germany (65%), and Australia (59%). Text generation via LLMs is the most common use (41% of firms), followed by data processing via machine learning and visual content creation. Adoption is higher at younger, larger, more productive, and higher-paying firms.
• Modest personal usage: Over two-thirds of executives personally use AI, but only at an average of 1.5 hours per week. CEOs use AI more frequently than CFOs or other executives. Usage has risen sharply since early 2025, roughly doubling from 0.9 to 1.4 hours/week among UK executives.
• Negligible realized impacts so far: More than 90% of executives report no effect of AI on employment over the past three years, and 89% report no impact on labor productivity (measured as sales per employee). Average realized productivity impact is only +0.29% and employment impact is essentially 0% across firms.
• Large expected future impacts: The same executives predict that AI will boost labor productivity by an average of +1.4% and output by +0.8% over the next three years, while cutting employment by -0.7% (equivalent to 1.75 million fewer jobs across the four countries by 2028). US executives are the most optimistic, predicting a +2.25% productivity gain.
• Expectations gap between employers and employees: Workers expect AI to increase employment by +0.5% at their firms over the next three years,  the opposite of what their employers predict, revealing a significant divergence in perceptions.

The authors document a striking “expectations paradox”: firms are broadly using AI, yet report near-zero impacts to date, while simultaneously anticipating large future productivity gains, a pattern consistent with diffusion lags in general-purpose technologies.

Firm Data on AI
Authors: Ivan Yotzov, Jose Maria Barrero, Nicholas Bloom, Philip Bunn, Steven J. Davis, Kevin M. Foster, Aaron Jalca, Brent H. Meyer, Paul Mizen, Michael A. Navarrete, Pawel Smietanka, Gregory Thwaites, Ben Zhe Wang
From: Bank of England, Instituto Tecnológico Autónomo de México (ITAM), Stanford University, Federal Reserve Bank of Atlanta, King’s College London, Deutsche Bundesbank, University of Nottingham, Macquarie University

A granular analysis of the AI “expectations paradox”

In this paper, Baslandze et al. survey 750 senior financial executives (mostly CFOs) at US firms via the long-running CFO Survey (jointly run by Duke University and the Federal Reserve Banks of Richmond and Atlanta), fielded during November–December 2025 with a supplementary wave through January 2026. Unlike Yotzov et al. above, which focuses on whether firms use AI and broad categorical impacts across four countries, this paper asks executives to quantify AI-attributed changes in revenue, employment, and capital, enabling a formal growth-accounting decomposition of labor productivity into capital deepening and residual TFP.

The two papers are notably convergent overall. However, there are a few areas of disagreement or incomplete alignment as far as the US in considered:

• Scale of expected productivity gains: Yotzov et al. report an average expected productivity boost of +1.4% over the next three years across four countries, while Baslandze et al. report CFOs expecting +3.0% in 2026 alone for the US. Even accounting for the different time horizons, the US-only CFO Survey implies a considerably more optimistic near-term outlook than the cross-country executive survey.
• Employment expectations: Yotzov et al. find executives expect AI to cut employment by -0.7% on average, whereas Baslandze et al. find a much smaller aggregate decline of less than -0.4% in 2026, with small firms actually expecting modest employment gains from AI.
• Nature of past productivity impacts: Yotzov et al. report that 89–90% of executives saw no impact on productivity or employment over the past three years, essentially a zero effect. Baslandze et al., by contrast, find that implied revenue-based productivity gains in 2025 were already positive (+0.6% overall, up to +0.8% in finance), suggesting some measurable effect has already materialized, even if modest. This is a meaningful difference in whether past effects are zero or already slightly positive.
• Role of capital deepening: Baslandze et al. explicitly decompose productivity gains and show capital deepening accounts for only 15% of the effect, with TFP driving the rest. Yotzov et al. do not perform this decomposition and as such, Baslandze et al. provide a more complete picture that implicitly challenges any interpretation that AI-linked investment (capital deepening) is what’s moving the needle.

Artificial Intelligence, Productivity, and the Workforce: Evidence from Corporate Executives
Authors: Salomé Baslandze, Zachary Edwards, John Graham, Ty McClure, Brent H. Meyer, Michael Sparks, Sonya R. Waddell, Daniel Weitz
From: Federal Reserve Bank of Atlanta, Federal Reserve Bank of Richmond, Duke University