At 10:00 a.m. Eastern on August 27, the Bureau of Labor Statistics released its Employment Projections for 2025 to 2035, filed as USDL-26-1422.
The opening line is flat. The U.S. economy is projected to add 5.9 million jobs over the decade, total employment rising from 170.3 million to 176.2 million, growth of 3.5 percent.
The second half of that sentence is the one to read. Over 2015 to 2025, the comparable figure was 10.9 percent.
The government now publishes an AI-exposure taxonomy
Boxed into the news release is a new product. Alongside the 2025-35 projections, BLS introduced a data set comparing occupations against one another on both their theoretical and their observed exposure to artificial intelligence, published on release day at bls.gov/emp/publications/ai-exposure-categories.htm.
The significance is not in the data. It is in the publisher.
Every AI-exposure ranking in circulation for the past two years has come from a consultancy, a hiring platform, or an academic paper with a position to defend. Each carries a business behind it. The national statistical agency has now issued its own classification. For the first time, an argument about whether AI displaces a given occupation has a public, non-commercial reference point on the table.
The group that loses most, and the reason BLS gives
Across the 22 major occupational groups, office and administrative support declines fastest: down 4.0 percent, shedding 752,100 jobs over the decade. It is both the steepest percentage decline and the largest absolute loss of any major group.
BLS does not hedge the mechanism. The continued integration of automation tools, including those powered by AI, into workflows is likely to reduce demand for several office and administrative support occupations.
Three other groups sit in the same paragraph. Sales and related falls 1.4 percent, as e-commerce keeps growing and AI moves into the sales process. Production falls 0.4 percent on continued deployment of automated machinery in manufacturing. For arts, design, entertainment, sports, and media, BLS writes that generative AI software, used to automate repetitive tasks and speed up processes, may limit demand for some jobs.
Four groups. Four mechanisms. One direction.
The fastest-growing occupations do not fill the hole
The growth side reads well until you count it.
Nurse practitioners post the fastest projected growth of any detailed occupation at 41.0 percent, adding 137,800 jobs. Data scientists grow 34.6 percent, adding 95,400. Computer and information research scientists grow 21.8 percent, adding 8,400. Solar photovoltaic installers grow 36.5 percent and wind turbine service technicians 29.5 percent, together adding fewer than 15,000 jobs across the whole decade.
Stack the AI-era winners by absolute count: 95,400 data scientists, 8,400 research scientists, under 15,000 in solar and wind combined. Under 120,000 in total.
One occupational group loses 752,100.
That gap does not close through retraining into those roles, because those roles are not large enough to absorb it. What actually absorbs the decade’s growth is elsewhere. Healthcare support grows 13.3 percent and healthcare practitioners and technical 8.0 percent, the two fastest-growing major groups, and between them they account for almost one-third of all new jobs created through 2035. Private healthcare and social assistance adds over 2.2 million jobs, about 37.0 percent of the decade’s total, and services for the elderly and persons with disabilities alone adds 625,400, the most of any detailed industry.
The typical new American job of the 2030s is not in a data center. It is in somebody’s living room or at a bedside.
Industries rise while occupations fall, and the two land on different people
The easiest thing to misread in this document is the gap between the industry view and the occupation view.
By industry, AI reads as a clean tailwind. Utilities is the fastest-growing major sector at 9.8 percent, nearly all of it in electric power generation, transmission and distribution, and BLS names AI power demand as part of that. Professional, scientific, and technical services is the third-fastest sector at 8.6 percent, adding 926,700 jobs on demand for AI systems, R&D, and related consulting. Computing infrastructure providers, data processing, web hosting and related services grows 25.1 percent, adding 120,400.
By occupation, the same force subtracts.
Both statements are true, and they land on different people. A professional services firm expanding because its clients want AI hires solution architects, data engineers, and delivery staff. The same firm shrinks its own administrative support, document processing, and scheduling. Sector headcount goes up. The people added and the people removed are not the same people and do not share a skill set.
We drew this line in August’s piece on the canaries: work that can be written down as a procedure goes first, tacit judgment goes last. The BLS outlook extends that line across the entire occupational classification.
The caveat BLS wrote itself
BLS boxes its own warning into the release. The projections give specific values, but that precision does not account for the inherent uncertainty of predicting long-term labor market change, and readers get more out of the direction and relative size of the projected changes than out of the point estimates. It adds that adjustments based on research into emerging technologies and new legislation are applied conservatively, generally requiring convincing evidence of long-term structural change.
Conservatively is the word to keep. The 752,100 figure is what survives after BLS deliberately holds the technology adjustment down.
Short-run evidence points the same way. In ADP’s August report on September 2, professional and business services lost 16,000 in a single month while education and health services added 45,000 — the ten-year structure already visible in one month of payroll records. Entry-level postings in law and consulting fell 35% over the year, and Stanford’s payroll study put employment for young workers in AI-exposed occupations down 13%.
For anyone making a career decision, this document is not a growth-rate leaderboard to memorize. It answers three questions. Is your major occupational group adding or shedding over the decade? How much of your actual work could be written down as a procedure? And if you move, does the destination BLS names require a credential, a license, or only tool fluency? Nurse practitioners grow 41 percent, and the entry ticket is a graduate degree plus licensure, not a weekend course.
