Why the yard still runs on people
The question “Will AI replace rail yard engineers?” gets a clearer answer once you split the shift into tasks. A yard engineer, dinkey operator or hostler spends the day moving switch engines and small industrial locomotives over short distances: pulling cuts of cars out of a track, shoving them into another, hauling locomotives to the fuel, sand and repair tracks, and tying things down at the end of a move. Almost none of that happens at a desk.
The work is also a conversation. The engine moves on hand signals and radio calls from the ground crew and the yardmaster. Somebody has to couple and uncouple cars, line switches, set handbrakes and watch the point while the engine shoves blind into a track. A rail yard is a changing place: cars in the wrong spot, a hose that will not seal, ice on a ladder, a contractor crew where nobody expected one. People improvise around that all day without writing it down.
The job is small and steady in size. About 3,920 people work as rail yard engineers, dinkey operators and hostlers in the US, with median pay of $60,600 and projected employment growth of 0.8% from 2025 to 2035 (BLS, 2025). That is not a profession being wound down. It is a narrow craft that each railroad, plant and quarry needs a handful of.
What AI does, what it helps with, and what stays with the crew
Software can take a slice of the task time on its own: 0%. The tasks that sit there are the record-keeping ones, like logging which cars moved where and generating the switch lists and shift reports that used to be written out by hand. A tool that files a movement report is not driving an engine.
A larger part of the job is work where AI assists a person: 13%. Yard-management systems already sequence moves and suggest the order of a switching plan, and machine vision reads car numbers and flags obvious defects as equipment rolls past a camera portal. Both give the engineer better information. Neither releases the brakes. Our coverage figure for this job, 10 out of 100, measures how much task time AI can handle today; you can read how that is built on the coverage scoring method.
The rest stays with people: 87% of task time. That is the hands-and-feet half of the job. Coupling and uncoupling cars, setting and releasing handbrakes, lining switches, checking air on a cut, and riding or protecting a shove move while reading signals in rain, heat and ice. Remote-control locomotive systems have already moved some of that away from the cab, but the person holding the control box is still a trained railroader standing in the yard.
How strong is the evidence?
Our evidence grade for this job is D. That is the grade we use when no published study has tested an AI system against a qualified yard engineer on this job’s own tasks, so we publish no parity number at all. Plenty has been written about automated mainline train control and about camera-based car inspection, but that is adjacent work, not a head-to-head test of yard switching.
What would settle it is specific and measurable. A trial in a working yard, reported by an operator or a regulator, comparing an automated or fully remote switching setup against a crewed engine on moves completed per shift, coupling and derailment errors, delays caused by handling exceptions, and safety events. Until something like that exists, the honest answer is that the yard half of the job has not been measured. The quality parity method explains how a graded result would change the page.
What could move the timeline
Most likely after 2044 (8 in 10 of our scenarios). For what that window is and is not, see the replacement-year method.
Two things could pull it earlier. Yards are enclosed, slow and mapped, which makes them friendlier to automation than a public road, and remote-control operation is already normal in many of them, so the control problem is partly solved. Second, the software side is cheap next to a wage, as the cost panel on this page shows. Once a capability works, the budget case is not the obstacle.
Two things hold it back. Most of what is left is physical: our robotics read puts this job’s ground work at the dexterous humanoid tier, meaning a machine would need hands, balance and footing on ballast and car ladders in bad weather. That hardware is not in service. Second, yard work sits inside federal safety rules, railroad operating practice and labor agreements that set who may move equipment and how. Those change slowly, and for good reason. Our wider look at humanoid robots and physical jobs covers why hands lag software.
How to stay needed around the engine
Lean into the parts of the shift that software is not close to. Coupling, air and handbrake work on live equipment. Protecting and directing shove moves where judgment about clearance and people on the ground decides whether anyone gets hurt. Spotting defects and odd behavior on cars and locomotives before they become a bad order or a derail.
Two skills pay off. Remote-control locomotive operation, because that is where the work is already drifting and the certifications are concrete. And reading data: switch lists, yard-management screens and inspection flags are the inputs a modern yard runs on, and the engineer who trusts them correctly, including knowing when to ignore one, is the one the yardmaster calls.
What to do: ask your railroad which remote-control and yard-management systems it is rolling out next, and get on the training list early.
Closely related work is scored the same way here: locomotive engineers, railroad brake, signal and switch operators, and railroad conductors and yardmasters. You can see the whole group on the rail transportation workers family page, the wider industry on our transportation and warehousing sector page, and where this job lands against hands-on work generally on the list of jobs that mostly need a person. To weigh two of these side by side, use the job comparison tool, and if you want the full scoring approach it is set out in our methodology.