SAFE
The job is walking platforms, rail yards and train cars, intervening in fights and medical emergencies, clearing trespassers off track right-of-way, and making arrests — none of which a language model or current robot can do. AI does compress the paperwork side: incident reports, video review, crime-pattern mapping and fare-evasion analytics are already being automated, and smart camera systems reduce the number of officers needed for pure surveillance posts. What remains is a sworn, personally liable human who must decide in seconds whether to use force and then defend that call in court.
Nearly all of this fall was the 2020 shock. It has been climbing back since.
Median pay $71,820 → $90,230 +0.5% 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
+3%
Percentage only. The projection counts a different population from the 4,390 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 +3% 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.
~200 openings a year on average, including replacing people who leave.
OfficerPatrollerPatrolmanPatrol ManTrack PatrolPatrol OfficerPolice CaptainTrack WatchmanRailroad PoliceTransit OfficerUnarmed OfficerPolice SpecialistRailroad WatchmanRailroad DetectiveTransit SpecialistField Training AgentField Training AdvisorTransit Police OfficerTransportation OfficerLaw Enforcement OfficerRailroad Police OfficerTransportation SergeantSecured Entrance MonitorTransit Authority Police
Holding it up: embodiment . Weakest point: trust premium .
Tasks largely resist digitisation Physically removing an unresponsive person from a rail car, sweeping a yard for trespassers on live third-rail track, and cuffing a resisting subject on a moving train are the core duties — AI takes the incident narrative, the CAD entry and the camera monitoring, which is why this sits at 15 rather than 19.
Hands-on in uncontrolled environments Shifts are spent on foot on platforms and ballast, in tunnels, on right-of-way with hostile weather, live traction current and moving equipment, plus grappling with people who do not want to be grappled with — an uncontrolled environment by any definition, held just below the ceiling only because some hours go to station posts and report writing.
Licensed human required and personally liable These are sworn peace officers under state POST certification with railroad police commissions (49 U.S.C. 28101 for interstate authority); the officer's own name is on the arrest affidavit and the use-of-force report, and a bad stop lands on them personally in a §1983 suit — the 14 rather than 18 reflects that the agency indemnifies and the badge, not an individual practice, generates the work.
Exists to be accountable for ambiguous calls Deciding in seconds whether a man on the platform edge is suicidal, intoxicated or diabetic, whether to draw, and whether to hold or release a juvenile trespasser are unscripted high-stakes calls; general orders and use-of-force continua constrain the range enough to keep this at 15 rather than 19.
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 (15/20 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 (14/20) is whether the law requires a licensed human to sign. Trust premium (11/20) is whether buyers specifically pay for a person. Judgment and accountability (15/20) 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 40 of this occupation's 73 points (55%).
Embodiment (18/20) 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 83/100, still SAFE.
State POST (peace officer standards and training) rules or transit-agency policy requiring that any AI-flagged detection — gun detection, fare-evasion analytics, facial recognition hit — be independently corroborated and the arrest decision signed by a sworn officer, with the officer named as declarant on the probable-cause affidavit. Already partially in motion: several states (e.g. Utah, Washington, Maryland facial-recognition statutes) bar FRT matches as sole basis for arrest and require human corroboration; extension to transit weapons-detection pilots (NYC MTA Evolv trial) is the checkable step.
Two-tier structure is real here: report writing, video review and pattern mapping are the routine tier; platform intervention and testimony are the judgment tier. If agencies formally reassign report-drafting and analytics to civilian analysts plus AI tools, the sworn role's remaining day is almost entirely non-automatable field work.
Task-mix shift: as camera-based monitoring absorbs static surveillance posts, the remaining deployment is response-only — use-of-force, crisis intervention, track-intrusion rescue. Formalized if a transit authority contract or FTA safety-management-system rule designates the officer as the accountable decision-maker for de-escalation and mental-health calls (Metro Transit and BART crisis-intervention specialist tiers are precedents).
Union contract language (e.g. Amtrak FOP, MTA PBA, SEPTA TWU-adjacent police units) requiring minimum sworn-officer staffing per station or train consist independent of camera coverage — bargained headcount floors that survive surveillance buildout.
The limit. Trust premium has no realistic route: riders do not choose their transit police and cannot pay for a human alternative, so the dimension is capped by the absence of a purchasing decision. Embodiment is already near ceiling at 18. The larger risk to this occupation is not AI capability but headcount: smart-camera justified staffing cuts and substitution of unsworn ambassadors/civilian analysts shrink the workforce even while each remaining job scores high.
| Philadelphia-Camden-Wilmington, PA-NJ-DE-MD | 80 | $103,980 +15% |
| Philadelphia-Camden-Wilmington, PA-NJ-DE-MD | 80 | $103,980 +15% |
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 73. 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.