Firms predict an AI productivity boom is coming

Comparable international survey data on artificial intelligence adoption by firms is still lacking. This column surveys over 5,000 CFOs, CEOs, and executives from stratified firm samples across the US, UK, Germany and Australia. It shows that around 70% of firms actively use AI, particularly younger, more productive firms. Firms report little impact of AI over the last three years, with over 80% of firms reporting no impact on either employment or productivity. At the same time, firms forecast AI will boost productivity by 1.4%, increase output by 0.8%, and cut employment by 0.7% over the next three years.

Interest in the impact of artificial intelligence (AI) on the economy has grown rapidly in the last few years. ¹ This trend has produced a large quantity of data from a variety of sources, including surveys, Census data, job postings, etc. (e.g. Acemoglu et al. 2022a, 2022b, Bonney et al. 2024, McKinsey 2025). These data often face challenges related to sample size, representativeness, and the nature of the responses. In many cases, data on firm-level AI use do not come from senior executives who can provide accurate responses. As a result, there is no single high-quality, large-sample representative international survey of AI use as reported by senior executives.

This column summarises results from a recent paper on AI adoption by businesses and the realised and expected impacts of AI technologies on firm employment and productivity (Yotzov et al. 2026). We use survey data from over 5,000 CFOs, CEOs, and executives across four economy-wide firm samples. US data are collected from the Survey of Business Uncertainty (SBU), run by the Federal Reserve Bank of Atlanta.  UK data are collected from the Decision Maker Panel (DMP), run by the Bank of England.  German data are collected from the Bundesbank Online Panel – Firms (BOP-F), run by the Deutsche Bundesbank.  Australian data are collected from the Business Outlook Scenarios Survey (BOSS), organised by Macquarie University.  To ensure consistency, all four surveys used the identical questions and were timed to run between November 2025 and January 2026.

The majority of businesses are currently using AI technologies

Across all four countries, we find that a majority of businesses are currently using some AI technology. The results on current use of AI technologies are summarised in Figure 1. Adoption is highest in the US (78% of firms), followed by the UK (71%), Germany (65%), and Australia (59%). On average, 69% of all firms are currently using AI. ² The most popular current use is ‘text generation using LLMs’ (by 41% of firms on average), but around 30% of firms also report using data processing using machine learning and visual content creation. ³   We also analyse how firms expect adoption of AI technologies to evolve. Across all four countries, 75% of firms expect to be using some AI technology over the next three years.

Figure 1 Current use of AI technologies by businesses

Notes: This figure is based on responses to the question: “Which of the following artificial intelligence technologies, if any, does your business currently use?” Firms could select more than one option. The data from the US Survey of Business Uncertainty was collected in November 2025. The data from the UK Decision Maker Panel was collected over November 2025 – January 2026. The data from the German Bundesbank Online Panel-Firms (BOP-F) was collected in January 2026. The data from the Australian Business Outlook Scenarios Survey was collected in December 2025. The results from the SBU, DMP, and BOP-F are employment-weighted; the results from the BOSS are unweighted. The results for all firms are the average of the four surveys, weighted by the respective number of responses.

AI adoption varies by multiple firm characteristics. These breakdowns use data from the DMP survey, due to the larger sample size and the availability of more firm covariates. More productive, larger, and higher-paying firms are more likely to be using AI. At the same time, firms with older directors on average, and older firms are less likely to be currently using AI. Use of AI technologies also varies by sector. We show that current AI adoption by UK firms is highest in the finance and insurance and professional and scientific sectors.

Firms expect AI will reduce employment over the next three years

Firms are asked to estimate the impact of AI adoption on their employment using five categories, ranging from a large positive impact (increasing employment by more than 5%) to a large negative impact (decreasing employment by more than 5%). We assign numerical values to each category to estimate the average impacts.

The employment impact of AI on average across all firms is essentially zero over the past three years. Indeed, more than 90% of businesses estimate no impact so far. However, there is heterogeneity across the four countries. In the UK, firms estimate that AI has lowered employment by around -0.14% over the last three years. In the US, there is also a small negative average impact of -0.09%.

Going forward, firms expect much larger impacts of AI on employment. 63% of businesses across the four surveys expect no impact, and 26% expect a negative impact on employment (Figure 2, Panel A). On average, firms across the four countries expect AI to lower employment by around 0.7% over the next three years, or around 0.23% per year (Figure 2, Panel B). The largest effects are in the UK (-1.4%), followed by the US (-1.2%). German and Australian firms do not expect AI to have as large an effect on overall employment over the next three years, possibly due to more regulated labour markets.

Figure 2 Expected impact of AI on employment over the next three years

A) Distribution of responses    

B) Average cumulative impacts

Notes: This figure is based on responses to the question: “How has the adoption of artificial intelligence technologies affected the number of employees in your business over the past three years? And how do you expect this to affect your number of employees over the next 3 years?” The impact for all firms is the average of the impacts for the four surveys, weighted by the respective number of responses. To calculate the average impacts (Panel B), values are assigned to each of the options in Panel A: large negative/large positive impacts are treated as ±7.5%; small negative/small positive impacts are treated as ±2.5%. 90% confidence intervals are shown for these impacts.

To gauge how lower employment may be achieved, a sub-sample of UK firms was asked a follow-up question about the expected importance of hiring fewer new employees versus increased exits of existing employees. ⁴   Around two-thirds of the reduction in employment is expected to come from firms hiring fewer new employees.

One important caveat is that this analysis refers to the expected employment impacts in existing firms. AI technologies may lead to the creation of new businesses and professions, which could change the net employment outcomes.

Firms also predict a sizeable boost to productivity from AI technologies

We next present results on the impact of AI technologies on firms’ realised and expected productivity. The survey questions follow the same structure as the employment impacts. Productivity in this analysis is defined as the volume of sales per employee.

Across all firms, AI is reported to have boosted productivity by around 0.29% over the past three years. Nevertheless, 89% of all firms report no impact of AI on their productivity, suggesting that the impacts are limited and contained in a small subset of businesses.

Firms expect much larger positive impacts of AI technologies on their productivity over the next three years. 60% of all firms expect no impact, and 37% of firms expect a positive boost to productivity (Figure 3, Panel A). On average, firms across the four countries expect AI to increase productivity by an average of around 1.4%, equivalent to around 0.5% per year (Figure 3, Panel B). The largest effects are in the US (+2.3%), followed by the UK (+1.9%), Australia (+0.9%), and Germany (+0.9%).

Figure 3 Expected impact of AI on productivity over the next three years

A) Distribution of responses   

B) Average cumulative impacts

Notes: This figure is based on responses to the question: “How has the adoption of artificial intelligence technologies affected the volume of sales per employee in your business over the past three years? And how do you expect this to affect your volume of sales per employee over the next 3 years?” The impact for all firms is the average of the impacts for the four surveys, weighted by the respective number of responses. To calculate the average impacts (Panel B), values are assigned to each of the options in Panel A: large negative/large positive impacts are treated as ±7.5%; small negative/small positive impacts are treated as ±2.5%. 90% confidence intervals are shown for these impacts.

Overall, the results show that the adoption of AI technologies has had little impact on firm employment and only a small positive impact on firm productivity so far. However, firms anticipate large impacts over the next few years. On average, businesses expect AI to boost productivity by around 1.4% over the next three years, while lowering employment by around 0.7% over the same period. This also implies an increase in output of around 0.8%.

There is a big gap in expectations between business executives and employees

In addition to the survey data of firm executives, in December 2025 we also surveyed US employees in the Survey of Working Arrangements and Attitudes.  We find that employees are far more optimistic than executives about the impact of AI on future employment, while less optimistic about the positive productivity impacts (Figure 4). Employees predict that AI will increase employment by approximately 0.5% in their firms over the next three years compared to the prediction from executives that it will reduce employment by 0.7% in all firms and 1.2% in US firms. Hence, there appears to be a large gap in the perceptions of the impact of AI, from a view by employees that AI will create jobs versus a view from executives that it will reduce jobs. Likewise, employees expect AI to increase productivity by around 0.9% over the next three years, compared to an expected increase of 1.4% across all firms and 2.3% by US firm executives.

Figure 4 Expected impacts of AI on employment and productivity by employees

Note: Responses to the questions: “How has the adoption of Artificial Intelligence technologies affected the NUMBER OF EMPLOYEES who work for your employer: over the past three years? over the next three years? How has the adoption of Artificial Intelligence technologies affected your employer’s SALES PER WORKER (PRODUCTIVITY): over the past three years? over the next three years?”
The sample includes respondents to the December 2025 SWAA wave who worked for pay during the week prior to the survey and pass our attention-check questions. We restrict attention to wage and salary employees (excluding self-employed workers and contractors). We reweight the raw responses to match the 2024 US population in cells defined by the cross product of age, sex, education and earnings.

Conclusion

In this column, we present new survey evidence from four large, economy-wide business surveys using identical questions asked between November 2025 and January 2026. AI technologies are currently used by around 70% of businesses, and adoption is expected to increase. Firms estimate that AI has had little impact on their employment so far, and only a modest boost to productivity over the past three years. However, firms expect AI to have much bigger impacts over the medium term. Over the next three years, firms predict that the adoption of AI will boost productivity by around 1.4%, on average. At the same time, businesses expect AI will reduce employment by around 0.7% over the next three years. We will continue to ask these questions to monitor the adoption of AI technologies and expected impacts as they evolve. We furthermore emphasise the importance of consistency in survey design and timing for obtaining comparable, high-quality results across countries.

Source : VOXeu