A new analysis shows hands-on American jobs—like rail track repair and water treatment—remain the hardest for artificial intelligence to replace.
Story Highlights
- Study lists specialized manual and machine-operator roles as least exposed to artificial intelligence.
- Jobs that mix physical work with person-to-person interaction stay resilient across studies.
- Global research says most work will change task by task, not disappear overnight.
- Clerical and language-heavy roles face higher exposure, while skilled trades look safer.
Study Identifies Jobs With Near-Zero Artificial Intelligence Exposure
Axios reported that a 2025 review placed several blue-collar and infrastructure roles near zero exposure to artificial intelligence. Examples included dredge operators, bridge and lock tenders, water treatment plant operators, foundry mold and core makers, and rail-track laying and maintenance equipment operators. These jobs require strength, spatial judgment, and on-site skill that current systems cannot match. The list underscores a simple point: when work depends on bodies, tools, and terrain, software alone cannot take it over.
That finding lines up with broader evidence on how artificial intelligence touches work. The International Labour Organization found about one in four workers have some exposure to generative artificial intelligence, but most roles will change rather than vanish. Workers keep the core duties while new tools handle narrow tasks. That pattern fits what Americans see on the ground: some office chores speed up, but jobs that keep water clean, rails straight, and streets safe still need people on site to get results.
Why Physical, Context-Rich Work Stays Hard To Automate
Anthropic’s researchers built a measure that tracks both what large language models can do and how people actually use them. They found higher exposure in desk tasks and customer messaging, not in jobs that hinge on touch, movement, and real-world sensing. When work mixes machine operation, teamwork, and changing environments, add-on software helps but rarely replaces the crew. That is why construction support, certain healthcare support, and field repair stay more secure than keyboard-only roles.
Policy groups and economists reach similar conclusions using the government’s job-task data. Studies that map job activities in the Occupational Information Network show that exposure rises with tasks computers can handle and falls with tasks that require situational judgment and manual control. The key is task mix, not job titles alone. If a role leans on hands, eyes, and coordination, models struggle. If it leans on text, numbers, or routine forms, models advance faster into the workflow.
Which Roles Face More Pressure From Automation
Clerical and language-centered positions face more risk as tools write drafts, summarize records, and route calls. A United Nations brief on generative artificial intelligence reports that clerical support workers carry the highest share of tasks with medium to high exposure. That does not mean a pink slip tomorrow, but it does mean faster change, re-training needs, and fewer entry roles if businesses shift routine work into software first.
Axios’ reporting points the same direction from the other side. It highlights customer service and other language-first jobs among the most vulnerable. As systems get better with long documents and chat, office-heavy roles feel the push first. That is why many companies are testing artificial intelligence for inboxes, knowledge bases, and scheduling. Meanwhile, the field jobs that keep power, water, freight, and bridges moving still depend on human crews who solve messy problems in real time.
What This Means For Workers, Families, And Policy
For American workers, the path is clear. Learn to use new tools where they help, but bet on skills that serve neighbors in the real world. Trades, public works, and maintenance have staying power because the work happens in dirt, steel, and concrete, not just in files. That aligns with a conservative view of dignity in work: strong families and strong towns need people who can fix what breaks and keep vital systems running when it counts.
For leaders, the priority is practical training, not lofty promises. Support apprenticeships, community colleges, and industry partnerships that turn willing Americans into qualified operators, linemen, mechanics, and technicians. Keep pathway jobs open by cutting red tape that blocks entry and by focusing federal efforts on real infrastructure, not trendy schemes. As artificial intelligence grows, the country should back the calloused hands that hold it all together—and protect the jobs that machines still cannot do.
Sources:
uscareerinstitute.edu, entrepreneur.com, forbes.com, arxiv.org, medium.com, bigissue.com
