SAFE
The job is welding cracked sills, replacing couplers, wheelsets and brake shoes, jacking cars, and crawling under equipment in yards and repair-in-place tracks — physical work in weather, on uneven ballast, with heavy irregular components no current robot handles. AI can absorb the paperwork tier: defect coding, AAR billing repair cards, parts ordering, and increasingly automated wayside detectors that flag hot bearings and wheel impacts before a human looks. What survives is the FRA-qualified person actually certifying air brake tests and single-car tests, and deciding whether a car gets patched, shopped, or bad-ordered out of service.
This fall is concentrated in 2020 and has not recovered since.
Median pay $56,390 → $67,530 -4.2% in real terms
This line is counted by the Bureau of Labor Statistics — the one figure on this page that isn't a judgement of ours. Headcount moves on demand, offshoring, demographics and the business cycle, and automation is one term among several, often not the loudest.
So a falling line is not evidence that AI did it, and a rising one is not evidence that it won't. Both happen in this register: some occupations resist automation and shrink anyway, others are highly automatable and keep growing. The marked year is 2020.
BLS projection, 2024–2034
+2.8%
Percentage only. The projection counts a different population from the 21,350 above — it includes self-employed workers, which for this occupation is most of them, so the two headcounts are not comparable.
Hard to automate, and growing
The work resists current AI and the BLS projects +2.8% more of these jobs by 2034. Note that safe does not mean well paid — several of the fastest-growing resistant occupations are among the lowest paid on the register.
Different clocks. The score is what current AI could do to this work today. The projection is how many of these jobs will exist in 2034. Everything between the two — how fast employers actually adopt, whether demand grows in the meantime — is why they can point opposite ways without either being wrong.
~1,500 openings a year on average, including replacing people who leave.
GearmanBrake LinerAir Brake ManBrake RelinerPlow MechanicRepair WorkerBrake AdjusterCoach MechanicRepair LaborerValve MechanicValve RepairerDrop Pit WorkerRail Car WelderRail SpecialistSignal MechanicTipple MechanicAir Brake RiggerAir Brake WorkerBrake SpecialistBreaker MechanicFreight RepairerInterior MechanicMine Car MechanicMine Car Repairer
Holding it up: embodiment . Weakest point: trust premium .
Tasks largely resist digitisation Cutting out a cracked center sill and welding in a new section, pressing wheels onto axles, changing a Type E coupler knuckle in a yard track, and rigging a 100-ton car onto jacks are single-piece, non-repeating operations on equipment that never presents the same way twice — the only genuinely automatable slice is the AAR job-code paperwork, which is why this sits at 17 rather than mid-band.
Hands-on in uncontrolled environments You work outdoors on ballast in whatever weather the yard gives you, crawl under cars behind blue-flag protection, torch and weld in confined body positions, and carry brake shoes, draft gear and slack adjusters by hand — the only thing keeping it off 20 is that some shop work happens indoors on a rip track with overhead cranes.
Licensed human required and personally liable 49 CFR 232 and 215 require you to be a designated qualified person to perform and certify Class I brake tests and single-car air brake tests, and your name on that record is what the FRA cites after a derailment — but the railroad or contract shop holds the certification program and the enforcement usually lands on the carrier, not on a personal state licence you could lose, which caps this at 12.
Meaningful discretion Deciding whether a car with a cracked sill or thin wheel flange gets a running repair, gets bad-ordered and shopped, or is safe to move empty to a repair point is a real call with derailment consequences, but AAR interchange rules, FRA condemning limits and wheel gauges give you hard numbers for most of it, which is why this is 12 and not 16.
The verdict above describes this occupation as a whole. Almost nobody does the typical version of a job — tick what's actually in your week and see how your own mix sits.
Your task mix speaks to task resistance (17/20 here) — how much of the day's work current AI already does. That is the dimension the boxes above are about.
It cannot move the other three. Liability shield (12/20) is whether the law requires a licensed human to sign. Trust premium (7/20) is whether buyers specifically pay for a person. Judgment and accountability (12/20) is whether the role exists to own consequential calls. Those are facts about the occupation's standing, not about which tasks are in your week — a paralegal who does only trial exhibits still holds no licence. Together they are 31 of this occupation's 67 points (46%).
Embodiment (19/20) is also a property of the work rather than the worker, but we don't tag individual tasks as physical or not, so the picker can't tell you anything about it. That's a limit of this tool, not a claim.
Did we get the list right? Tell us what's missing — the tasks are written from the outside, and you're reading this from the inside.
The moves above are yours to make. This is the other half: what would have to change in the world for the occupation itself to score higher. None of it is in any one person's gift, but it is where the floor actually comes from. Scores here are not a one-way ratchet. Only two of the five dimensions — task resistance and embodiment — track what machines can do. The other three track law, what buyers will pay for, and who is answerable, and those move in both directions, often in response to the same pressure AI creates. If every lever below landed, this occupation would score around 81/100, still SAFE.
Bad-order/shop-or-patch calls becoming the explicit locus of accountability as wayside detectors multiply: if AAR Field Manual rules or carrier rules require a qualified person to document rationale for overriding or accepting a detector flag (hot bearing, wheel impact load detector alarm), the role owns the consequential call under ambiguity rather than executing a queue.
Task-mix shift as detector networks and AI absorb defect coding, AAR billing repair cards and parts ordering: the residual day becomes non-routine structural welding, jacking, and disputed-defect adjudication. Already visible in carrier mechanical departments where car inspectors' clerical load has moved to systems while the wrench-and-certify tier stayed.
FRA tightening 49 CFR Part 215 / Part 232 qualification rules so that a named, individually-qualified inspector must personally certify Class I air brake tests and single-car tests, with the certification non-delegable to automated wayside or machine-vision inspection output — the reverse of current railroad petitions for waivers allowing automated track/car inspection (ATIP-style) to substitute for walking inspections. A rulemaking that explicitly bars automated inspection from satisfying the certifying-person requirement would raise this.
Post-derailment legislation in the mold of the Railway Safety Act of 2023 (S.576, introduced after East Palestine) enacting minimum mechanical-inspection time per car and requiring a qualified mechanical inspector — not a train crew member or a detector — to sign off, with recordkeeping tied to the individual's qualification number.
Union agreements (e.g. TCU/BRC carmen) or FRA whistleblower enforcement establishing that a carman's refusal to release a car cannot be overridden by a mechanical desk or algorithmic dwell-time target — making the individual's call binding, as pressed in carmen's testimony on reduced inspection staffing.
The limit. No realistic route to a higher trust premium: the buyer is a railroad or leasing company optimizing car-hire cost, and no shipper pays extra for human-performed repairs. Embodiment is near ceiling already and cannot rise. Total headroom is roughly 8-10 points, concentrated in liability_shield, and it is contested — carriers are actively petitioning FRA in the opposite direction to substitute automated inspection for human walking inspections.
| New York-Newark-Jersey City, NY-NJ | 800 | $92,250 +37% |
| Houston-Pasadena-The Woodlands, TX | 790 | $59,310 -12% |
| Chicago-Naperville-Elgin, IL-IN | 570 | $62,250 -8% |
| Dallas-Fort Worth-Arlington, TX | 370 | $59,820 -11% |
| Washington-Arlington-Alexandria, DC-VA-MD-WV | 250 | $103,540 +53% |
| Baltimore-Columbia-Towson, MD | 230 | $89,430 +32% |
| Denver-Aurora-Centennial, CO | 200 | — |
| Beaumont-Port Arthur, TX | 190 | $56,820 -16% |
| Washington-Arlington-Alexandria, DC-VA-MD-WV | 250 | $103,540 +53% |
| New York-Newark-Jersey City, NY-NJ | 800 | $92,250 +37% |
| Baltimore-Columbia-Towson, MD | 230 | $89,430 +32% |
We have no reported case of a named organisation automating this occupation. Not one deployment, not one announcement.
That is worth saying out loud next to a score of 67. The verdict above is about what the work exposes — what current AI could do to these tasks. It is not a claim that anyone has done it. For this occupation those two things have come apart completely: the capability argument is on this page, and the evidence column is empty.
Has AI actually changed your work? One tap, anonymous, and the running tally is public. Nothing else is asked of you.
Rather than check back: get the digest and we'll tell you what changed — or watch a single occupation from its own page.