SAFE
The adjudicative core — weighing credibility of live witnesses, ruling on objections in real time, sentencing, and owning a decision that deprives people of liberty or property — is constitutionally and statutorily reserved for a human officeholder with personal accountability. What AI already does well is the surrounding paper: summarizing briefs and records, checking citations, drafting routine orders and boilerplate findings, and triaging high-volume dockets like traffic, small claims, and warrant applications. Expect clerk-level drafting support to compress, which shifts the judge's day further toward hearings and contested rulings, not away from the bench.
Most of this decline happened after 2021 — it is not the pandemic dip.
Median pay $136,910 → $153,990 -10.0% 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
+2.5% 27,300 → 28,000 on the projections basis
Hard to automate, and growing
The work resists current AI and the BLS projects +2.5% 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.
~900 openings a year on average, including replacing people who leave.
JudgeJuristJusticeMagistrateChief JudgeCounty JudgePolice JudgeTribal JudgeCircuit JudgeLegal RefereeProbate JudgeTrial JusticeCriminal JudgeDistrict JudgeElection JudgePolice JusticeChancery MasterMunicipal JudgePresiding JudgeBankruptcy JudgeMagistrate JudgeImmigration JudgePolice MagistrateTrial Court Judge
Holding it up: liability shield . Weakest point: embodiment .
Tasks largely resist digitisation Reading a witness's hesitation on cross, ruling on a hearsay objection in the two seconds before the answer lands, and calibrating a sentence to a specific defendant's record and remorse are not text-prediction problems — but the 14 rather than 18 reflects how much of a judge's week is bench memos, scheduling orders, uncontested defaults, warrant and probable-cause forms, and boilerplate findings of fact that a model can draft to signature-ready quality.
Some physical or field component You have to be bodily present in the courtroom to keep order, hold a party in contempt, and be seen doing it — plus jail and video arraignments, in-camera inspections, occasional jury views of a premises, and for magistrates the after-hours on-call warrant duty — yet all of it happens in a controlled, secured building with a bailiff, which is why this sits at 10 and not with the field-based trades.
Licensed human required and personally liable The office itself is the shield: judges are elected or appointed under state constitutions or Article III, admitted to the bar, bound by a code of judicial conduct enforced by a commission on judicial performance, and no order has legal effect until a commissioned human officer signs it — an unsigned AI-generated ruling is void, not merely unwise.
Exists to be accountable for ambiguous calls Setting bail, terminating parental rights, granting or denying a suppression motion that decides the case, and imposing a sentence within a wide statutory range are decisions made on incomplete records, reviewable only for abuse of discretion, and attributed by name in a published opinion that binds future parties.
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 (14/20 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 (20/20) is whether the law requires a licensed human to sign. Trust premium (16/20) is whether buyers specifically pay for a person. Judgment and accountability (20/20) 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 56 of this occupation's 80 points (70%).
Embodiment (10/20) 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 89/100, still SAFE.
Task-mix shift as AI absorbs the paper tier: if bench-book drafting, citation checking, and routine order generation are handled by court-adopted tools (e.g., the AI tools already piloted in the 11th Circuit and several state administrative offices), the residual day is live hearings, credibility findings, contested evidentiary rulings, and sentencing — the tier AI cannot do at usable quality. Judicial officer headcount is set by caseload statutes and judgeship bills, not by per-case labor hours, so compression of the paper tier does not shrink the role.
If appellate reversals or due-process rulings establish that algorithmic input into liberty decisions is constitutionally suspect — extending State v. Loomis-style scrutiny of risk-assessment tools from advisory to prohibitive — litigants' demand for a human decider becomes an enforceable entitlement rather than a preference. Watch state supreme court rules on AI in adjudication (Illinois, Delaware, Texas have issued policies).
If rules like the standing orders now issued by dozens of federal district judges (post-Mata v. Avianca) requiring certification of AI use in filings are extended to the court's own work product — barring generative drafting of findings of fact or sentencing rationales — the drafting tier stays human by rule rather than by capability.
Private arbitration is the one competitive market for adjudication; if institutional rules (AAA, JAMS) formally bar AI arbitrators or require party consent, the paid-for-a-human signal is explicit in a priced market.
The limit. Three of five dimensions are at or near ceiling; realistic total headroom is roughly 6-8 points, capped by embodiment (a courtroom is a predictable indoor environment and remote hearings are now routine, so embodiment has no upward route and may fall).
| New York-Newark-Jersey City, NY-NJ | 1,280 | $204,160 +33% |
| Austin-Round Rock-San Marcos, TX | 710 | $153,990 +0% |
| Atlanta-Sandy Springs-Roswell, GA | 550 | $144,790 -6% |
| Seattle-Tacoma-Bellevue, WA | 550 | $135,180 -12% |
| San Juan-Bayamon-Caguas, PR | 490 | — |
| Houston-Pasadena-The Woodlands, TX | 480 | $135,430 -12% |
| Cleveland, OH | 440 | $80,020 -48% |
| Dallas-Fort Worth-Arlington, TX | 430 | — |
| Providence-Warwick, RI-MA | 150 | $229,760 +49% |
| Longview-Kelso, WA | 40 | $228,330 +48% |
| Boston-Cambridge-Newton, MA-NH | 420 | $221,710 +44% |
NPR reports that courts in China have ruled in favour of workers dismissed due to AI-driven automation, amid continued job displacement anxiety.
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.