September 2, 2026 09:00 AM PST
(PenniesToSave.com) – The Bureau of Labor Statistics put two documents into the world on August 27, 2026, and most of the coverage that followed treated them as one. The first was the agency’s employment outlook for 2025 to 2035, which projects that the American economy will add 5.9 million jobs, a gain of 3.5 percent, taking total employment from 170.3 million to 176.2 million [3]. Growth over the coming decade is expected to run slower than it did over the last one [3]. The second document was new. It sorts every detailed occupation the agency tracks into four tiers of relative exposure to artificial intelligence [7].
Headlines merged the two into a single story about AI erasing American jobs. The agency’s own documentation says something narrower. The exposure tiers are not a forecast of employment growth or decline, not a wage forecast, and not an estimate of how likely any occupation is to be automated [7]. That distinction matters for any household weighing what to train for, what to fall back on, and how much of the next ten years to plan around a federal table.
Quick Links
- Which Jobs Does the Labor Department Expect to Shrink Fastest?
- Which Occupations Stand to Lose the Most Actual Jobs?
- What Does the Government’s New AI Exposure Ranking Actually Measure?
- Why Does the Bureau Say Its AI Rankings Do Not Predict Job Loss?
- Where Is the Labor Department Projecting Growth Instead?
- How Much Weight Should a Ten Year Projection Carry?
Which Jobs Does the Labor Department Expect to Shrink Fastest?
The steepest projected decline by percentage belongs to word processors and typists, an occupation the agency expects to fall 34.4 percent, from 40,400 workers in 2025 to about 26,500 in 2035, against a 2025 median wage of $49,280 [4]. Telephone operators follow at a 27.6 percent decline, dropping from roughly 3,500 to 2,500 at a $41,740 median [4]. Switchboard operators are projected down 26 percent, from 35,400 to 26,200 at $38,630, and data entry keyers down 25.5 percent, from 131,800 to 98,200 at $41,340 [3]. Rounding out the top ten are foundry mold and coremakers at 23 percent, patternmakers in metal and plastic at 22.8 percent, telemarketers at 21.4 percent, hand grinding and polishing workers at 19.2 percent, mining roof bolters at 18.9 percent, and hand cutters and trimmers at 18.9 percent [4].
The next ten continue the pattern. Print binding and finishing workers and order clerks are each projected down 17.5 percent, model makers in metal and plastic down 17.3 percent, forging machine setters down 17.2 percent, engine and other machine assemblers down 17 percent, carpet installers down 16.4 percent, payroll and timekeeping clerks down 15.9 percent, file clerks down 15.8 percent, underground mining loading and moving machine operators down 15.8 percent, and pressers for textile and garment materials down 15.7 percent [4].
Most of these occupations sit below a $50,000 median, with telemarketers, sewing machine operators, and pressers among the lowest paid on the list [2]. That is worth pairing with a hard look at how to build a household budget that can absorb a wage interruption. The pattern is not universal, though. Roof bolters earn a $78,540 median and underground loading operators earn $74,500, both well paid skilled trades appearing on the same decline list [4]. Scale also deserves a caveat. Telephone operators lose roughly 1,000 positions over ten years only because about 3,500 people still hold the job at all [3]. The Hill’s account describes thirty declining occupations, but the full thirty and the figures below rank twenty are not reproduced in the reporting reviewed here [2].
A 34 percent decline in an occupation of 40,400 people is a smaller event than a 6.5 percent decline in an occupation of 3.1 million.
Which Occupations Stand to Lose the Most Actual Jobs?
The rate list and the volume list are two different lists, and the volume list is where most American families will find themselves. Cashiers lead it with a projected loss of 200,600 positions, falling from more than 3.1 million workers in 2025 to about 2.9 million by 2035, which the agency scores as a 6.5 percent decline [2]. Office clerks follow at 156,200 positions lost, customer service representatives at 141,800, secretaries and administrative assistants at 114,100, and bookkeeping, accounting, and auditing clerks at 85,600 [3]. Shipping, receiving, and inventory clerks give up 62,800 positions, first line supervisors of retail sales workers 52,500, tellers 44,700, data entry keyers 33,600, and correctional officers and jailers 29,500 [3].
Put the two lists side by side and the story shifts. Word processors post the steepest rate in the country and shed about 13,900 jobs. Cashiers post a modest 6.5 percent and shed roughly fourteen times as many [2]. Taken as a whole, office and administrative support is projected to fall 4 percent and shed 752,100 jobs, the largest decline of any major occupational group [6]. Sales occupations, along with farming, fishing, and forestry, are also expected to lose ground [3], and federal government employment is projected down 3.4 percent [4].
Breyon Williams, chief labor market economist at Groundwork Collaborative, told CBS News that the picture is not one directional, saying “there are some gains from AI demand coupled with losses” in predictable places [4]. Groundwork Collaborative is an advocacy organization, and his framing should be read with that in mind. What the numbers describe is plain enough on their own. The jobs most exposed by raw headcount are ordinary retail, clerical, and customer facing work, the same positions that have long served as entry points, second incomes, and bridges between careers for American households. A decade is enough time to prepare, which is the argument for treating an automated emergency fund as infrastructure rather than an aspiration.
What Does the Government’s New AI Exposure Ranking Actually Measure?
The second document released on August 27 is the one driving most of the alarm, and it is worth understanding on its own terms. The Bureau built it to complement the projections, grouping every detailed occupation into four tiers of relative exposure labeled low, moderate, high, and very high [7]. Relative is the operative word. The tiers compare occupations to one another rather than describing any absolute level of risk [7].
The ranking draws on five outside data sources, three theoretical and two based on observed usage [7]. On the theoretical side, Felten, Raj, and Seamans asked survey respondents whether specific AI applications relate to abilities required at work, using 2020 occupational data. Eloundou, Manning, Mishkin, and Rock tested whether large language model capabilities could cut the time needed for occupational tasks, drawing on expert raters and GPT-4 in 2023. Eisfeldt, Schubert, Taska, and Zhang applied a related framework using GPT 3.5 Turbo the same year [7].
The two observed measures come from private companies. Massenkoff and McCrory at Anthropic built an exposure measure from Claude.ai conversations and programming interface traffic mapped to occupational tasks, weighted toward tasks with higher automative than augmentative use [7]. Tomlinson, Jaffe, Wang, Counts, and Suri at Microsoft measured Copilot activity matched to intermediate work activities [7]. The agency then converted each source to percentile ranks on a zero to one scale, mapped everything to the 2018 Standard Occupational Classification, and used a clustering algorithm to assign the four tiers [7]. TechInformed reports that 206 occupations landed in the highest tier, including web developers and customer service representatives, though that count does not appear in the agency documentation reviewed here [6].
Two private companies’ product usage data is now an input to a federal career planning tool. That is worth saying out loud.
Why Does the Bureau Say Its AI Rankings Do Not Predict Job Loss?
Because the Bureau said so directly, in writing, on the same page that publishes the tiers. “Exposure does not imply job loss, productivity gains, automation probability, or wage effects,” the agency states [7]. Its listed interpretation limits go further. An exposure category is not a forecast of employment growth or decline. High or very high exposure does not necessarily mean employment will fall. Low exposure does not mean an occupation is safe from future technological change. It is not a wage forecast, an adoption probability, a worker replacement estimate, or a productivity forecast, and it does not distinguish AI that replaces work from AI that assists it [7]. The agency also notes that the observed measures do not directly confirm whether anyone in a given occupation used AI on the job [7].
The agency’s own numbers demonstrate the point. Web developers carry very high exposure and are projected to grow nearly 4 percent. Customer service representatives carry very high exposure and are projected to fall 5 percent, or 141,800 jobs [6]. Same tier, opposite outcomes. Nor does the Bureau credit AI alone for the steepest clerical declines. Its occupational utilization analysis points to automated phone systems for telephone and switchboard work, a combination of substitution by other workers and machine learning for word processing, and a mix of AI, machine learning, mobile data capture, and optical character recognition for data entry [6].
The measurement limits received almost no coverage and deserve some. The theoretical sources reflect AI capabilities available no later than mid 2023. The observed sources may skew toward early adopters of particular models. Most of the inputs focus on language modeling and exclude image and video generation. Results produced by language models depend on the specific model and prompt used. None of the five sources account for bottlenecks or complementarities inside an occupation. And of 4,155 possible occupation and source combinations, 211 values were imputed rather than observed, affecting 75 occupations [7].
Where Is the Labor Department Projecting Growth Instead?
Health care and social assistance carries the decade. The sector is projected to grow 9.5 percent and add 2.2 million jobs, roughly 37 percent of all new American jobs over the period, driven by an aging population and rising rates of chronic conditions including heart disease, cancer, and diabetes [4]. Health care occupations account for half of the twenty fastest growing jobs [4]. The agency’s own summary of the release leads with the same material, highlighting the fastest growing occupations overall and those requiring postsecondary education rather than the declines that drove the coverage [1].
At the occupational level, nurse practitioners lead at 41 percent growth against a $132,300 median, followed by solar photovoltaic installers at 37 percent and $53,140, data scientists at 35 percent and $120,230, wind turbine service technicians at 30 percent and $64,120, and medical and health services managers at 24 percent and $123,860 [4]. Physical therapist assistants grow 23 percent at $68,380, information security analysts 21 percent at $129,180, and industrial machinery mechanics 18 percent at $64,520 [4]. Fast growing does not always mean well paid. Home health and personal care aides appear on the same list at 18 percent growth with a $35,800 median [4]. Computer and mathematical occupations are projected up 7 percent, roughly double the rate for all occupations [6].
Energy figures need their aggregation named. At the sector level, CBS News reports utilities as the fastest growing industry at 9.8 percent while adding only 58,800 jobs [4]. At the detailed industry level, Staffing Industry Analysts reports solar electric power generation at 152.9 percent and wind at 62.1 percent, though solar, wind, geothermal, and other power generation combined add just 35,800 positions, against 625,400 for services for the elderly and persons with disabilities [5]. There is a policy tension here worth naming rather than smoothing. The Trump administration has moved to curtail renewable energy projects, yet the projections still show growth in the segment, and Vestas chief executive Henrik Andersen told analysts the United States is concluding it needs more of everything [5]. Professional, scientific, and technical services are projected to add nearly 927,000 jobs tied to AI demand [4].
How Much Weight Should a Ten Year Projection Carry?
Less than the headlines imply, and the Labor Department is the first to say so. The agency warns that its data cannot account for the inherent uncertainty of predicting long term changes in the labor market [2]. It acknowledges separately that any employment effects from artificial intelligence are highly uncertain, while arguing the information still helps with career planning [7]. Projections are revised on an annual cycle, and the 2026 to 2036 set is due in 2027 [7].
For a household, the two lists answer different questions. A rate of decline tells you about an occupation’s trajectory. A total job change tells you how crowded the competition for remaining openings is likely to get. Neither is a promise, and neither describes a specific employer in a specific county. The occupations on the decline list still employ substantial numbers of Americans today, including 131,800 data entry keyers and more than 3.1 million cashiers [3].
Read any of this as one input among several, weighed against local labor conditions, actual posted wages, and the real cost of any retraining. A federal table published in Washington does not know what the manufacturers, hospitals, and retailers in your county intend to do. The decision about what to train for belongs to the worker and the family, and the agency’s own caveats support exactly that reading.
Final Thoughts
What the August 27 release documents is clear enough. Clerical, phone based, and hand craft occupations are projected to decline fastest by rate. Retail and administrative work is projected to lose the most positions by count. Health care and skilled technical work carry nearly all of the growth, and the pay range inside that growth is wide.
What the release does not document is that artificial intelligence is the cause, that the exposure tiers predict outcomes, or that any of it is settled. The Bureau said as much in its own limitations section, and the gap between what was published and how it was reported is the part worth remembering the next time a ten year forecast arrives with a headline attached. The two figures a family can act on right now are ordinary ones: the median pay in the occupations projected to shrink, and how many people still hold those jobs today. Both point toward the same unglamorous work of building margin, which is what a handful of practical money saving steps are ultimately for.
Works Cited
[1] U.S. Bureau of Labor Statistics. “BLS Occupational Employment Projections 2025-35.” U.S. Bureau of Labor Statistics, 27 Aug. 2026, www.bls.gov/video/home.htm?video=FFYVz2Ghsgs.
[2] Bink, Addy. “These 30 Occupations Could See the Fastest Job Decline in the Next Decade: Labor Department.” The Hill, 1 Sept. 2026, thehill.com/business/economy/6060815-these-30-industries-could-see-the-fastest-job-decline-in-the-next-decade-labor-department/.
[3] Walrath-Holdridge, Mary. “US Jobs Most Likely to Grow or Shrink in the Next Decade. See List.” USA Today, 1 Sept. 2026, www.usatoday.com/story/money/economy/jobs-labor/2026/09/01/us-jobs-projected-decline-growth-next-decade/91566834007/.
[4] Cerullo, Megan. “See Which Jobs Are Forecast to Grow Fastest and Slowest Over the Next Decade.” CBS News, 1 Sept. 2026, www.cbsnews.com/news/fastest-growing-jobs-labor-department-projections/.
[5] Johnson, Craig. “Solar, Wind Power Generators Expected to See AI-Fueled Job Surge.” Staffing Industry Analysts, 28 Aug. 2026, www.staffingindustry.com/news/global-daily-news/solar-wind-power-generators-expected-to-see-ai-fueled-job-surge.
[6] Gupta, Aman. “Bureau of Labor Statistics Adds over 200 Occupations in Top AI-Exposure Tier.” TechInformed, 1 Sept. 2026, techinformed.com/bureau-of-labor-statistics-adds-over-200-occupations-in-top-ai-exposure-tier/.
[7] U.S. Bureau of Labor Statistics. “Artificial Intelligence (AI) Exposure Categories.” Employment Projections, U.S. Bureau of Labor Statistics, 27 Aug. 2026, www.bls.gov/emp/publications/ai-exposure-categories.htm.