AI, Jobs & the Economy: what the data says

Last reviewed Oct 3, 2026. Statistics and projections are labeled as such; opinions are not included.

How much AI is used at work

Three independent measures point the same way: use is rising, and it is still a minority of workers and firms in the United States.

• Workers. In a Pew Research Center survey of 5,010 U.S. workers in September 2025, 21% said at least some of their work is done with AI, up from 16% a year earlier. 65% said they do not use AI much or at all in their job. Use was higher for workers with a bachelor’s degree or more (28%) than for those with some college or less (16%).¹
• Businesses. The Census Bureau’s Business Trends and Outlook Survey found overall AI use between 17% and 20% from December 2025 to May 2026. Use was 37% among firms with at least 250 employees, and under 20% among firms with four or fewer. By sector, Information (39.7%) and Finance and Insurance (33.9%) were above the national rate of 19.8%; Retail Trade was around 14%.²
• Larger organizations worldwide. The AI Index 2026 reports that 88% of surveyed organizations used AI in at least one business function in 2025, generative AI was used in at least one function at 70%, and deployment of AI agents was in the single digits across nearly all functions.³

The Census wording changed in November 2025 from AI “in producing goods or services” to “in any business function”, so earlier BTOS figures are not directly comparable.²

Is AI taking jobs yet?

The short answer from the main sources is: not in the overall totals so far, with early signs of change for the youngest workers in some exposed occupations. The sources disagree about how much weight to put on those signs.

• The AI Index 2026 says large-scale job losses have not yet shown up in overall employment data, but that employment of U.S. software developers aged 22 to 25 has fallen nearly 20% since 2024 while older developers’ headcount grew. It also reports that one-third of surveyed organizations expect AI to reduce their workforce in the coming year, most often in service operations, supply chain and software engineering, and almost half expect little or no change.^3,4
• The OECD’s Employment Outlook 2026 calls the evidence “mixed” and mostly U.S.-based. Some studies find no significant employment effects of generative AI; others find a disproportionate effect on early-career workers. The OECD adds that some results may reflect a weaker economy rather than AI, because AI-exposed occupations are concentrated in sectors that cut hiring quickly when conditions worsen. Its data from Australia, Canada, the EU and the United States suggest that AI’s role in the unemployment gap for young entrants “remains limited”.⁵
• The International AI Safety Report 2026 says early evidence shows no effect on overall employment, with some signs of declining demand for early-career workers in AI-exposed occupations such as writing, and that economists disagree on the size of future effects: some expect losses to be offset by new jobs, others expect widespread automation could reduce employment and wages.⁶
• The BLS says uncertainty about AI’s impact on its 10-year projections “remains very high”. It adjusts projections only where there is convincing evidence, and says most adjusted changes are concentrated in computer and mathematical occupations, arts, design, entertainment, sports and media, various business administration occupations, and legal work.⁷

What forecasts say

Forecasts are scenarios built on assumptions. They are useful for the direction and for what could go wrong, not as predictions of what will happen.

• World Economic Forum (employer survey, 1,000+ companies). By 2030, the WEF projects 170 million new roles created and 92 million displaced, a net gain of 78 million, with job disruption equal to 22% of jobs. These figures cover all major trends (technology, demographics, the green transition, the economy), not AI alone. Among employers, 77% plan to upskill workers, and 41% plan to reduce their workforce as AI automates tasks. The WEF lists graphic designers among roles now declining as generative AI reshapes work, and care, education and frontline jobs among the largest in growth.⁸
• International Monetary Fund. The IMF estimates that about 60% of jobs in advanced economies may be impacted by AI, against 40% in emerging markets and 26% in low-income countries. For advanced economies it says roughly half of the exposed jobs may benefit from AI integration, while for the other half AI may do key tasks now done by people, which could lower labor demand, wages and hiring. “Exposed” does not mean “lost”.⁹
• OECD, by region. The OECD reports that the share of workers in jobs exposed to generative AI ranges from about 16% in some regions to more than 70% in others, depending on local industries. It also cites an estimate that AI could raise annual productivity growth by 0.25 to 0.6 percentage points over the next 10 years.⁵

Productivity

The AI Index 2026 summarizes studies reporting productivity gains of 14% to 15% in customer support, 26% in software development and 50% in marketing output. It says gains are largest in structured, measurable work and smaller where deeper reasoning is needed, and notes recent evidence that heavy reliance on AI may carry long-term learning penalties.³

The International AI Safety Report adds a caution: some studies find developers using AI assistants finish certain tasks 20–30% faster, but one study found experienced programmers were 19% slower on complex tasks even though they believed they were faster. It describes the evidence on how AI augments human work as inconclusive.⁶

On the company side, the AI Index reports that consumers’ estimated value from generative AI in the United States reached $172 billion a year by early 2026, and that AI companies’ revenue is rising quickly while compute and infrastructure spending are also at record levels.³

The sources compared

Reading the numbers

• Exposure is not loss. A job is “exposed” when AI could do some of its tasks. The IMF and OECD figures measure exposure, and both say exposure can help or hurt.
• Projection is not measurement. WEF, BLS and the IMF describe what could happen under assumptions. Pew, Census and the AI Index’s employment data describe what has been measured so far.
• Surveys ask people what they do or expect. Pew counts workers’ own reports; AI Index expectations of layoffs are what organizations say they expect, not what happened.
• Where the data is from matters. Most of the youth-employment evidence is from the United States, and the OECD says so.⁵

Keep reading

• AI Rules & jobs (laws, plus a shorter jobs summary)
• AI by Country
• AI Timeline
• How AI works: plain-English guides
• News