The U.S. Bureau of Labor Statistics released its 2025–2035 employment projections on August 27, 2026, and paired them with four relative AI exposure categories for every detailed occupation it tracks. In the same release, BLS said the U.S. economy is projected to add 5.9 million jobs from 2025 to 2035.
Alongside that broad forecast, BLS said demand for AI-based systems, research and development, and associated consulting services is expected to drive 926,700 new jobs in professional, scientific, and technical services over the projection period. The release also says employment in office and administrative support occupations is projected to fall by 752,100 jobs over the decade. Axios, in coverage published on August 28, 2026, described the release as a more nuanced picture than a simple AI job apocalypse narrative and reported that AI and automation are expected to reduce office and administrative support jobs by 752,100 over the next decade.
BLS says the new exposure categories are “low,” “moderate,” “high,” and “very high.” In a separate explainer released the same day, BLS said it used five external data sources to build the categories and said they are not forecasts of employment change, wages, or worker replacement. That scope matters: the categories mark relative exposure, not a prediction that any one occupation will gain or lose jobs, pay, or workers by a fixed amount.
What BLS added
The central change in the federal projections release is the addition of AI exposure categories to the official occupational outlook. BLS says the categories are relative and apply to every detailed occupation it tracks. The agency’s explainer says the four labels are “low,” “moderate,” “high,” and “very high.”
BLS also says the categories are not forecasts of employment growth or decline, not wage forecasts, and not estimates of worker replacement. That limitation is important because the projections release and the exposure explainer are doing different jobs: one gives a decade labor-market outlook, while the other classifies occupational exposure to AI using external data sources.
The supplied sources do not identify the full list of detailed occupations or explain how each occupation was categorized in this research bundle. They also do not state that the new exposure labels are intended to be a universal ranking of risk across the labor market; the sources only say BLS created four relative categories and paired them with its projections.
Where the projected gains are concentrated
In the release, BLS says demand for AI-based systems, research and development, and associated consulting services is expected to drive 926,700 new jobs in professional, scientific, and technical services over the projection period. Axios reported that figure in its August 28 coverage and used it to frame the release as more nuanced than a straightforward AI job apocalypse story.
The BLS release also says the economy overall is projected to add 5.9 million jobs from 2025 to 2035. The supplied sources do not identify which occupations inside professional, scientific, and technical services are expected to account for that projected increase, beyond the broad reference to AI-based systems, research and development, and associated consulting services.
Separate from that sector-level figure, the BLS release says projected growth in utilities is linked to data-center-related electricity demand, and it says the decade outlook includes a projected decline in federal government and retail trade employment. Those points are part of the same official release, but the supplied sources do not provide the underlying occupational breakdown for this article.
Where the projected losses show up
The clearest decline signal in the supplied sources is office and administrative support occupations. BLS says employment in that group is projected to fall by 752,100 jobs over the decade. Axios reported that AI and automation are expected to reduce office and administrative support jobs by that amount over the next decade.
That statement should be read narrowly. The BLS exposure categories are not a direct forecast that AI will remove exactly 752,100 jobs from any single occupation or subcategory, and the BLS explainer does not say the categories measure worker replacement. The sources also do not identify which office and administrative support occupations are most exposed, or whether any specific job within that group is expected to fall faster than the broader category.
More generally, the supplied sources support a mixed picture: the same release that pairs projected gains in AI-linked professional services with projected losses in office and administrative support also says the exposure categories are not employment forecasts. That means the categories should not be treated as a one-to-one map of winners and losers.
What this means for workers, employers, and planners
Verified reporting: BLS has now attached four AI exposure categories to its official projections, and those categories sit beside a decade outlook that includes both projected growth and projected decline. The supplied sources say the categories are relative, built from five external data sources, and explicitly not forecasts of employment change, wages, or worker replacement.
Analysis, with limits: That combination gives schools, employers, and policymakers a new official reference point for discussing AI exposure in the labor market, but the research bundle does not state how any of those groups will use it. It also does not establish whether the new taxonomy will change hiring, training, or investment decisions; it only shows that BLS has made AI exposure a measurable category in its projections.
What to watch next: The supplied sources do not identify any follow-up BLS releases, additional occupational detail, or revisions to the exposure categories. If future BLS materials break down the categories by occupation or sector, that would help show whether the labels line up with the jobs the agency expects to expand or contract. For now, the official message is narrower: AI exposure is now part of the federal projection framework, but the exposure labels themselves are not forecasts.
Source scope and wording
The supplied sources vary slightly in emphasis. The BLS projections release states the 5.9 million-job figure, the new exposure categories, and the decade outlook for sectors including professional services, utilities, federal government, and retail trade. The BLS explainer provides the category definitions and says the categories are not forecasts of employment change, wages, or worker replacement. Axios adds the 926,700-job and 752,100-job figures in its coverage and characterizes the release as more nuanced than a simple AI job apocalypse narrative.
Because those are different source functions, this article keeps their claims separate. The projections release gives the labor-market outlook; the explainer defines the categories; Axios supplies the interpretive frame and the office-support figure attributed to AI and automation. The supplied sources do not provide any stronger claim than that.
Sources
- Employment Projections and Occupational Outlook Handbook News Release – 2025 A01 Results — U.S. Bureau of Labor Statistics
- Artificial Intelligence (AI) exposure categories — U.S. Bureau of Labor Statistics
- Here are the top-growing jobs in the country — Axios