EXPOSED
A large share of the day is documentation — presentence investigation reports, case notes, chronologies, violation reports — and actuarial risk scoring (LSI-R, COMPAS) that is already algorithmic and drafts well from structured inputs. What does not automate is the field half: home and workplace visits, drug testing, arresting or detaining violators, sitting with a defendant's family, and testifying in court. The occupation is also a state function with civil-service and often union protection, which slows headcount cuts even as per-officer caseloads absorb the productivity gain.
Roughly flat across the period, with year-to-year wobble.
Median pay $54,290 → $66,270 -2.3% 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
+2.6% 92,300 → 94,800 on the projections basis
Growing, and only partly exposed
The BLS expects +2.6% more of these jobs by 2034, and at 58/100 the work is only partly exposed — some tasks are automatable, the core of the job is not. Nothing here is in tension.
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.
~7,900 openings a year on average, including replacing people who leave.
Court WorkerParole AgentPeople OfficerTruant OfficerProbation AgentDetention WorkerJuvenile OfficerProbation WorkerDetention OfficerParole SpecialistProbation OfficerAttendance OfficerJuvenile CounselorDetention AttendantJuvenile SpecialistParole Officer (PO)Probation CounselorCorrectional OfficerProbation SpecialistCorrections CounselorCorrectional CounselorCorrections SpecialistAdult Probation OfficerCorrectional Specialist
Holding it up: judgment & accountability . Weakest point: liability shield .
Mixed — a routine tier and a judgment tier Roughly half your week — PSI narratives, chronological case notes, violation petitions, condition-compliance checklists, and the LSI-R/COMPAS scoring itself — runs off structured intake data that a model drafts competently, which pulls this well below 14; the field-contact half (unannounced home visits, observed urinalysis, hands-on searches, court testimony) is what keeps it out of the 0-6 band.
Some physical or field component You drive to residences and job sites in neighborhoods you don't control, collect observed urine samples, sometimes carry and effect arrests on a bench warrant, and transport offenders — real uncontrolled-environment work, but it's bracketed by courthouse hearings and desk hours on the case management system, so it lands at 12 rather than the 16+ of a patrol deputy who is outdoors the entire shift.
Certification preferred, not legally required There's no state practice licence you personally hold at risk; entry usually requires a bachelor's plus a state academy and POST-style certification, and revocation risk runs through your employing agency and civil-service discipline rather than a board that can end your career independently — certification-preferred territory, not licensed-professional immunity.
Exists to be accountable for ambiguous calls You decide whether a positive test becomes a warning, a treatment referral, or a revocation petition that sends someone back to prison, and you write the sentencing recommendation the court usually follows — the override authority over the actuarial score is explicitly yours, and the consequences of getting it wrong appear in the newspaper, which is what puts this at 15 despite the structure that departmental violation matrices impose.
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 (10/20 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 (9/20) is whether the law requires a licensed human to sign. Trust premium (12/20) 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 36 of this occupation's 58 points (62%).
Embodiment (12/20) 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 77/100 — SAFE.
Genuine two-tier occupation: chronologies, case notes, PSI narrative assembly and actuarial scoring are the routine tier. If those are absorbed, the residual job is the tier AI cannot do at usable quality — motivational interviewing, cross-examinable court testimony, absconder location, family and employer negotiation, and the discretionary revocation call. Task-mix shift alone raises this without any new law, though it comes with caseload increases rather than headcount protection.
Classification of recidivism/needs risk assessment as a high-risk or prohibited-without-human-review use — the EU AI Act already lists risk assessment of natural persons for reoffending as high-risk, and Colorado SB 24-205 style state AI acts extend duties to consequential government decisions. A US state adopting a mandatory documented human override step on any AI-generated supervision-level or violation recommendation converts a drafting tool into something a named officer must adjudicate.
A state statute or judicial-branch rule requiring that presentence investigation reports and probation revocation recommendations be authored and signed by a sworn officer, with AI-generated content disclosed on the face of the document — the pattern already visible in court AI-disclosure orders (federal district judges' standing orders on generative AI, Texas 5th Circuit rulemaking) and in state pretrial algorithm statutes (e.g. Idaho HB 118 requiring transparency and barring sole reliance on risk tools). If the signature is attributable and challengeable on appeal, the human is structurally load-bearing.
Extension of the Wisconsin v. Loomis holding (risk score may not be determinative) into an affirmative rule that the officer must state independent reasons for a supervision or revocation recommendation, reviewable at hearing. Also raised if state legislatures narrow qualified immunity or create a duty-to-supervise cause of action after a high-profile failure-to-revoke case, making the discretionary call personally consequential.
Contractual or statutory minimum face-to-face contact standards — American Probation and Parole Association contact-standard adoption, or a union contract (AFSCME, SEIU state corrections locals) specifying minimum in-person home/employment visits per supervision level and capping caseload per officer. Court consent decrees following supervision-failure litigation have imposed similar contact floors. This converts field work from discretionary to mandated.
Shift of drug testing, GPS/EM installation and compliance verification, and warrant service in-house rather than vendor-outsourced, as some states have done after monitoring-vendor scandals. Each is a physical task in an unpredictable environment that must be performed by the supervising agency.
The limit. No plausible route on trust_premium. The buyer is the state, not the supervised person; there is no market in which a defendant or a court pays extra for a human probation officer, and civil-service structure means what protects headcount is union and statute, not preference. Treat the current 12 as the ceiling, not a floor to build on.
| Los Angeles-Long Beach-Anaheim, CA | 4,700 | $108,150 +63% |
| New York-Newark-Jersey City, NY-NJ | 3,490 | $79,370 +20% |
| Houston-Pasadena-The Woodlands, TX | 2,320 | $51,820 -22% |
| Philadelphia-Camden-Wilmington, PA-NJ-DE-MD | 1,890 | $74,310 +12% |
| Chicago-Naperville-Elgin, IL-IN | 1,560 | $67,610 +2% |
| Dallas-Fort Worth-Arlington, TX | 1,340 | $59,170 -11% |
| Riverside-San Bernardino-Ontario, CA | 1,260 | $96,970 +46% |
| Sacramento-Roseville-Folsom, CA | 1,250 | $100,690 +52% |
| San Jose-Sunnyvale-Santa Clara, CA | 560 | $144,410 +118% |
| San Francisco-Oakland-Fremont, CA | 830 | $135,890 +105% |
| Los Angeles-Long Beach-Anaheim, CA | 4,700 | $108,150 +63% |
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 58. 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.