A major international study examining AI’s impact at the firm level has delivered a surprisingly measured conclusion: the disruption so many predicted has not yet materialised at scale. Instead, the data points to incremental shifts in productivity and employment—signals of early-stage deployment rather than technological underperformance.
The working paper, published by the National Bureau of Economic Research and produced in collaboration with researchers from the Federal Reserve Bank of Atlanta, Bank of England, Deutsche Bundesbank, and Macquarie University, surveyed nearly 6,000 verified executives across four countries.
Its headline finding: more than 90% of firms report no measurable change in headcount attributable to AI over the past three years.
Early adoption, limited aggregate impact
At first glance, the results may appear underwhelming. Yet historically, general-purpose technologies—from electricity to the internet—tend to diffuse gradually before reshaping entire sectors. AI appears to be following a similar trajectory.
Adoption itself is widespread. Roughly 69% of firms report using some form of AI. Large language model–based text generation leads at 41%, followed by machine learning for data processing at 28%, and visual content creation at 29%. In the UK, firm-level adoption has risen steadily.
AI tools are increasingly embedded in day-to-day workflows. The lag lies not in deployment, but in translating usage into measurable firm-level outcomes.
Executives anticipate acceleration
While recent impacts remain modest, executive expectations suggest stronger gains ahead.
On average, leaders project a 1.4% rise in productivity and a 0.8% increase in output over the next three years. US executives are more bullish, forecasting productivity gains above 2%, while UK firms expect slightly lower—but still meaningful—improvements.
In economies that have struggled with stagnant productivity growth for more than a decade, gains of this magnitude matter. Small percentage increases, when compounded across sectors, can materially shift national output.
Employment shifts will likely be gradual
Employment expectations are more nuanced. Executives anticipate a modest 0.7% reduction in headcount across surveyed countries. In the UK, much of this adjustment is expected to occur through slower hiring rather than direct layoffs. This pattern suggests gradual role reallocation rather than abrupt workforce contraction.
As with previous waves of automation, headline job figures rarely capture the full picture. AI deployment is simultaneously generating new categories of work—data governance specialists, model oversight professionals, AI implementation leads, and prompt engineers—roles that scarcely existed a few years ago.
The expectation gap between leaders and workers
The study also highlights a divergence between executive forecasts and employee sentiment.
Parallel survey data from US workers shows employees expecting a slight increase in employment at their firms, while executives anticipate mild contraction. Workers also project smaller productivity gains than leadership teams do.
This difference likely reflects perspective. Executives view AI through the lens of cost structures and competitive positioning. Employees experience AI more directly as task augmentation—tools that enhance performance rather than replace roles outright.
Evidence from controlled deployments of large language models in customer support and professional services supports this view. Productivity gains tend to concentrate among less experienced staff, often accompanied by improvements in quality. Where training and communication are clear, resistance to adoption appears limited.
Interpreting adoption data carefully
AI adoption figures vary significantly across surveys, underscoring how methodology shapes perception.
Executive-level studies often capture enterprise-wide deployments and strategic intent. Broader business surveys may report narrower usage definitions or earlier-stage experimentation. Sampling methods, respondent seniority, and question framing all influence reported adoption rates.
In this case, respondents were phone-verified, unpaid, and predominantly senior leaders such as CEOs and CFOs. The data was also cross-checked against macroeconomic employment and output figures from national statistics agencies, strengthening its credibility.
From experimentation to integration
The broader takeaway is not that AI has failed to deliver impact—but that it remains early in its integration cycle.
Executives appear to view the next three years as an inflection point. As systems mature, workflows stabilize, and integration improves, the gap between deployment and measurable economic outcomes may narrow.
The core question is no longer whether AI will influence productivity and employment. It is how effectively organisations can convert widespread adoption into sustained, quantifiable gains.
For now, optimism is cautious—but increasingly grounded in operational reality rather than hype.
Source: https://www.artificialintelligence-news.com/news/ai-impact-executives-optimism-for-the-future/
