In official job projections, artificial intelligence is one more technology that changes work gradually, and the numbers contain no AI shock. Federal labor statisticians (the Bureau of Labor Statistics, or BLS) released their 2025 to 2035 projections on August 27, 2026. Total employment is expected to grow 3.5 percent over the decade, from 170.3 million to 176.2 million jobs, with productivity and technological progress assumed to follow historical patterns. BLS describes the uncertainty about AI's effects as very high, which is why the figures work as a baseline and not as a prediction of what AI will do.
Key takeaways
- BLS projects 3.5 percent total job growth for 2025 to 2035, slower than the 10.9 percent recorded for 2015 to 2025.
- The projections assume technological progress in line with history. BLS adjusts an occupation only when research gives a clear direction and size for the effect.
- Office and administrative support is projected to lose 752,100 jobs (4.0 percent), and BLS names automation tools, including AI-powered ones, among the reasons.
- Data scientists are projected to grow 34.6 percent, and computer and information research scientists 21.8 percent.
- Canada's federal projection system covers 2024 to 2033, with a 2026 to 2035 update scheduled to appear in stages from fall 2026.
What the 2025 to 2035 projections assume about technology
A projection needs an assumption about how fast output per worker rises. BLS states that it assumes labor productivity and technological progress will match historical experience. Its release explains the reasoning: recent AI developments raise the prospect of faster productivity growth, but if that faster growth were uniform across all industries, employment would not change, and the agency lacks the data to model differences between industries. BLS therefore publishes one scenario built on historical patterns, and it points out that past technologies changed occupations gradually.
The same document says developments in AI are moving quickly and that uncertainty about their effects remains very high. The headline numbers answer a narrower question: what happens to jobs if the pace of change looks like the past.
When BLS does adjust for AI
A February 2025 article in the Monthly Labor Review by BLS economists Christine Machovec, Michael Rieley, and Emily Rolen describes the rule. The projection models are built to capture gradual structural change, and they are changed for a technology only when research gives clear expectations about the direction and size of its employment effect. The article gives two precedents. Employment of photographic process workers was cut in the projections and fell from 86,300 in 2004 to 9,200 in 2023. Truck driver employment was not adjusted for autonomous vehicles, and it grew from about 1.7 million in 2012 to over 2.2 million in 2023.
The authors also state that the timing and scale of generative AI effects are too uncertain to reflect fully, and that their case studies reflect information available in June 2024.
Occupations where AI appears in the explanation
The August 2026 release links AI to both growth and decline. On the decline side, it says the continued integration of automation tools, including those powered by AI, is likely to reduce demand for several office and administrative support occupations. That major group is projected to fall 4.0 percent, the fastest decline of any group, and to shed 752,100 jobs, the most of any group. Sales and related occupations are projected to fall 1.4 percent and production occupations 0.4 percent.
On the growth side, the release credits AI adoption and demand for related infrastructure, such as data centers, with supporting employment in computing infrastructure providers, and it cites rising electricity demand that includes AI power needs. Data scientists are projected to grow 34.6 percent and computer and information research scientists 21.8 percent. Nurse practitioners lead all detailed occupations at 41.0 percent.
Case studies from the previous projection cycle
The Monthly Labor Review article applied the method to specific occupations for the 2023 to 2033 cycle, when all occupations were projected to grow 4.0 percent. These figures are superseded by the 2025 to 2035 cycle but show how the same method gives different answers for jobs that use AI differently.
| Occupation | Jobs in 2023 (thousands) | Projected change, 2023 to 2033 |
|---|---|---|
| Software developers | 1,692.1 | +17.9 percent |
| Personal financial advisors | 321.0 | +17.1 percent |
| Paralegals and legal assistants | 366.2 | +1.2 percent |
| Credit analysts | 73.7 | -3.9 percent |
| Claims adjusters, examiners, and investigators | 345.2 | -4.4 percent |
The article adds that medical transcriptionists (-4.7 percent) and customer service representatives (-5.0 percent) are examples of occupations whose core tasks generative AI can most easily reproduce. For software developers it reports no clear evidence that AI productivity gains will outweigh growth in demand.
The Canadian counterpart
Employment and Social Development Canada runs the Canadian Occupational Projection System (COPS), which identifies occupations that may face labor shortage or surplus over a medium-term horizon. The current published set covers 2024 to 2033, and the program's page says an update for 2026 to 2035 will be released progressively starting in fall 2026. As of October 27, 2026 the program page still lists 2024 to 2033 as the latest set, with the next update in preparation. For 2024 to 2033, more than 100 occupations are expected to have a moderate or strong risk of shortage. The program page and its open data record do not state an AI-specific adjustment, so Canadian projections cannot be read as an AI forecast either.
Checklist for reading any job projection
- Which years does it cover, and which cycle replaced it?
- What does it assume about productivity and technology?
- Is the figure a change in jobs, a share of jobs, or annual openings?
- Does the source say AI explains the change, or does the reader infer it?
- Does the same source publish exposure data, and does it warn that exposure is not a forecast?
FAQ
Does the BLS projection predict AI job losses?
No. It projects employment under a technology-as-usual scenario and names AI as one reason demand may fall in some office and sales occupations. It also publishes AI exposure categories for 831 occupations, and it says that placing an occupation in an exposure category does not forecast whether its employment will rise or fall.
Why would a highly exposed occupation still grow?
Exposure measures how many tasks AI could do or assist, while a projection weighs demand for the occupation's output. The bureau also cautions that the categories cannot tell automation apart from augmentation.
How often are projections revised?
The BLS exposure page states that the 2026 to 2036 projections will be released in 2027, so the 2025 to 2035 set, released August 27, 2026, is the latest until then.
Is there a single Canadian number for AI and jobs?
Not in the federal projection system reviewed here. COPS reports openings and seekers by occupation without an AI adjustment on its program pages.
Related reading
- What unemployment data show about AI so far
- Whether AI can replace human workers
- Scientists and AI: tasks versus jobs
- AI tools in human resource management
More in AI and Society.
Last verified: October 2026
Sources
- Employment Projections: 2025-2035, news release USDL-26-1422, U.S. Bureau of Labor Statistics, August 27, 2026 (source page)
- Incorporating AI impacts in BLS employment projections: occupational case studies, Machovec C, Rieley M, Rolen E, Monthly Labor Review, U.S. Bureau of Labor Statistics, February 2025 (source page)
- Artificial Intelligence (AI) Exposure Categories, U.S. Bureau of Labor Statistics, page modified August 27, 2026
- Canadian Occupational Projection System (COPS), Employment and Social Development Canada, 2024 to 2033 projections, modified January 30, 2025