Why the work stays with road crews
Will AI replace highway maintenance workers? The honest answer sits in the work itself. Most of the day is spent patching pavement, setting cones and signs, flagging traffic, clearing debris, and fixing guardrails and drainage after a storm. Each job happens in a different place, in weather nobody chose, a few feet from moving vehicles. Software can tell a crew where a pothole is. Someone still has to stand in the lane, cut the edges, fill it and compact it.
The second reason is judgment on site. A shoulder that looks solid on camera can be soft underneath. A crash scene changes the plan for the rest of the shift. Crews read the road, the traffic and each other, and they adjust the taper and the closure as they go. That is coordination work, not data work, and it carries real safety stakes for the public and for the crew.
Pay and headcount give some context. The Bureau of Labor Statistics counted about 154,960 highway maintenance workers in the United States, with median pay near $50,260 a year, and projects employment growth of roughly 3.4% between 2025 and 2035 (BLS, 2025). That is steady demand tied to roads that keep aging, not a shrinking occupation. How that feeds the headline figure is explained on our methodology page.
What software handles, what it assists, and what people keep
Inspection and paperwork are where automation has the clearest hold. Vehicle-mounted cameras and machine vision now flag cracking, rutting and potholes from survey footage, and condition data can be turned into ranked work orders without a person reading every log. Share of task time where AI can do the work: 0%.
Assistance shows up in planning and equipment control. Models help rank which segments to patch first, time winter salt and plow routes against forecasts, and keep paving and grading machines on grade through automated controls. The operator is still there; the machine is taking over the fine corrections. Share of task time in the assisted group: 0%. The way we count a task as covered is set out under coverage scoring.
Everything physical stays with people. Patching and sealing pavement, erecting and repairing signs, barriers and guardrails, mowing and brush clearing on the roadside, snow and ice removal, and traffic control around a live closure all need hands and eyes on site. Share of task time that needs a human: 100%.
How strong the evidence is
There is no direct head-to-head test of AI against highway maintenance workers on this job’s core tasks. Our evidence grade reflects that: D. Because the grade is at that level, we publish no parity number for this occupation, and nobody should read one into the score.
What would settle it is specific. A measured trial of automated defect detection against trained inspectors on the same road miles, with false positives and misses counted. A field test of a machine completing a pothole repair to a state DOT specification, start to finish, including setup and cleanup. Documented work-zone traffic control run without a human flagger across a full season, with safety outcomes reported. Until results like those exist, this page rests on the task mix rather than on a benchmark. The same test applies to similar trades, which is why AI and trades careers tends to be a story about tools, not replacement.
When this could change
Most likely after 2048 (8 in 10 of our scenarios). What that window measures, and how we build it, is described under the replacement-year method.
Two things could pull the date in. First, inspection keeps getting cheaper: mounted cameras and drones already cut the time crews spend driving routes to look for damage, which shifts hours from finding problems to fixing them. Second, machine control on paving, grading and mowing equipment keeps spreading, so one operator covers more ground per shift. Neither removes the crew. Both can change how many people a district needs per mile.
Two things hold it back. Almost all of this job is physical, and the robotics tier it would take to do it is a dexterous humanoid working outdoors on uneven ground beside live traffic. Nothing at that level is available to buy and run at state-DOT scale. Cost is the other brake: software inspection is cheap, but a machine rugged enough for winter road work, plus the per-hour cost shown in the costs panel above, is a much harder purchase than hiring. Public procurement cycles and work-zone safety rules slow adoption further. The same pattern shows up across humanoid robots and physical jobs.
How to stay needed on the crew
Lean into the tasks the task list above leaves with people. Traffic control and work-zone setup is the clearest one: flagging, closures and tapers carry certification, legal weight and real risk. Pavement repair to specification is the second, from crack sealing to patching that holds through a freeze-thaw cycle. Structure and roadside repair is the third, including guardrail, signs, culverts and drainage after storm damage.
Two skills raise your value fastest. One is equipment and machine-control literacy: running plows, mowers, loaders and automated grade systems, and knowing when to override them. The other is inspection and reporting with digital tools, including drone and camera data, so you can check what the software flagged and write the work order it feeds.
What to do: ask your district who reviews the automated condition reports, and get on that list.
Nearby work is worth a look if you want to move sideways. Rail track laying and maintenance equipment operators sit closest on the data, with paving, surfacing and tamping equipment operators and construction laborers next. You can put any two of them side by side with the job comparison tool, see the wider group on the other construction and related workers family page, or read how the whole construction sector scores. If you want the broader picture, the list of jobs that mostly need a person shows where road work sits among them.