Across the United States, the jobs that kept office buildings humming for decades are starting to disappear. File clerks, data-entry specialists, bookkeepers, payroll processors: the Bureau of Labor Statistics projects that employment in office and administrative support occupations will shrink by roughly 4% between 2024 and 2034, eliminating hundreds of thousands of positions. The culprit is not offshoring or a recession. It is artificial intelligence, which can now handle the repetitive, screen-based tasks that once sustained an entire tier of the American workforce.
At the same time, a very different corner of the labor market is expanding. The BLS projects employment for electricians to grow 11% over the same period, well above the national average. Plumbers, pipefitters, and steamfitters are expected to see about 6% growth. Both fields pay a median salary above $61,000 a year, according to BLS data, roughly $20,000 more than the median for office and administrative support roles. The gap between these two trajectories is becoming one of the defining features of the American labor market in 2026.
Why office roles are shrinking
The decline is not hypothetical. A National Bureau of Economic Research working paper (No. 33509), authored by economists Daron Acemoglu and David Autor among others, constructed task-level AI exposure measures using data spanning 2010 to 2023. The researchers found that higher AI exposure within firms directly depresses labor demand, meaning companies that adopt AI tools hire fewer people for the roles those tools can perform. The effect is sharpest for jobs built around repetitive, easily codified tasks, exactly the kind of work that defines most office and administrative support positions.
The NBER study also found a partial buffer: when workers concentrate on a narrower, harder-to-automate set of duties, the labor demand losses shrink. That finding matters because it suggests the damage is not a blanket wipeout. Roles that blend administrative work with judgment, relationship management, or specialized knowledge may survive longer than pure data-processing jobs. But for the millions of Americans whose daily work consists largely of entering, sorting, or verifying information on a screen, the outlook is grim.
The International Monetary Fund reinforced these findings in Staff Discussion Note 2024/001, which examined how generative AI exposure falls unevenly across the workforce. Clerical and administrative roles, often held by women without advanced degrees, face especially high exposure. The IMF outlined two competing forces: augmentation, where AI makes workers more productive and potentially more valuable, and displacement, where it replaces them outright. The note warned that the balance between those forces will shape whether AI narrows or widens income inequality in the years ahead.
Why trades keep growing
Electricians and plumbers occupy the opposite end of the automation spectrum. Their work is physical, site-specific, and unpredictable in ways that current AI systems cannot replicate. No large language model can snake a clogged sewer line or wire a breaker panel in a building under construction. That fundamental constraint is one reason the BLS projects sustained growth for these occupations.
Federal policy is amplifying the trend. The Infrastructure Investment and Jobs Act, signed in 2021, authorized more than $1.2 trillion in spending, with significant portions directed toward broadband expansion, electric vehicle charging networks, power grid upgrades, and water infrastructure. The Inflation Reduction Act, passed in 2022, added billions more in clean energy incentives that require licensed electricians for installation and maintenance. As of spring 2026, much of that funding is still flowing into active projects, sustaining demand for skilled tradespeople in construction and maintenance.
An aging workforce compounds the pressure. The median age of electricians and plumbers skews older than many other occupations, and retirements are creating openings faster than training pipelines can fill them. The result is a seller’s market for anyone with a journeyman’s license and a willingness to work on-site.
The gap no one has closed
The obvious question is whether displaced office workers can realistically move into the trades. The answer is complicated. A typical electrical apprenticeship runs four to five years and combines paid on-the-job training with classroom instruction. Plumbing apprenticeships follow a similar timeline. For a 45-year-old bookkeeper facing a layoff, that is a significant commitment, even if the long-term earnings are better.
Neither the BLS projections nor the NBER research addresses retraining pathways with program-level data. The IMF note acknowledges the need for policy intervention but operates at a level of abstraction that offers little guidance to an individual worker trying to decide what to do next. Community colleges and union apprenticeship programs exist in most states, but capacity, geographic access, and financial support during training vary enormously.
Geographic variation adds another layer of complexity. National projections can mask sharp local differences. A region experiencing a construction boom tied to data center development or renewable energy installation may have far more demand for electricians than the national average suggests. Meanwhile, a metro area with a heavy concentration of back-office operations, think insurance processing hubs or corporate shared-services centers, may feel AI-driven job losses sooner and more acutely than the country as a whole. None of the major federal or academic sources offer city-level forecasts granular enough to guide individual decisions.
What the evidence actually supports
The strongest data points in this story come from two places. BLS occupational projections carry the weight of the federal government’s statistical apparatus and are updated on a regular cycle. They reflect modeled expectations rather than guarantees, but they represent the most authoritative baseline available for U.S. labor market trends. The NBER working paper offers empirical, task-level measurement of how AI exposure affects firm hiring, grounded in observed data rather than employer surveys or sentiment polls.
Together, these sources establish a clear directional signal: AI is suppressing demand for office work while leaving hands-on trades largely untouched. The IMF analysis adds a distributional lens, highlighting who bears the cost, but its conclusions about augmentation versus displacement are analytical frameworks, not measured outcomes. Readers should treat them as guides to possible scenarios.
Some widely cited reporting from outlets like Bloomberg and the Financial Times, drawing on World Economic Forum employer surveys, has flagged concerns that AI is closing the door on entry-level office opportunities. Those accounts align with the academic evidence but rely partly on what companies say they plan to do, not what has already happened. They are useful context, not proof of losses already locked in.
Where the labor market is headed
For anyone weighing a career decision in 2026, the practical signal is hard to ignore. Office and administrative roles face documented headwinds from AI over the next decade. Skilled trades tied to construction, maintenance, and electrification show stronger outlooks, backed by federal data, infrastructure spending, and the simple reality that physical work resists automation in ways that screen-based work does not.
None of this means every electrician or plumber will enjoy uninterrupted demand, or that every office job will vanish. Projections will be revised as AI capabilities evolve and infrastructure timelines shift. But the dividing line in the American labor market is becoming harder to miss: it runs between work that can be done on a screen and work that must be done with hands, tools, and physical presence. For workers on the wrong side of that line, the time to start planning is not next year. It is now.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
