SAFE
The core of this job is sitting in a room (or on a video call) with two hostile parties, reading credibility and body language, and moving them toward a deal they will actually sign — that is persuasion under live ambiguity, not text processing. AI already drafts settlement agreements, summarizes case files, models damages ranges, and runs low-stakes online claim resolution, so the paperwork tier and small-money consumer disputes are exposed. What holds is that arbitration awards are legally enforceable instruments issued by a named neutral on a court or AAA/JAMS roster, and parties pay specifically for a human whose neutrality and reputation they trust to bind them.
Dipped in 2020, then grew past where it started.
Median pay $63,930 → $75,530 -5.5% 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
+4.3% 9,100 → 9,500 on the projections basis
Hard to automate, and growing
The work resists current AI and the BLS projects +4.3% 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.
~300 openings a year on average, including replacing people who leave.
ArbiterMediatorOmbudsmanArbitratorAdjudicatorConciliatorLabor MediatorLegal MediatorFamily MediatorDivorce MediatorFederal MediatorLabor ArbitratorArbitration ManagerContract NegotiatorDispute CoordinatorArbitration SpecialistMediation CommissionerResolution CoordinatorLong Term Care OmbudsmanDebt Settlement NegotiatorPublic Employment MediatorEnvironmental Conflict ManagerPeacebuilding and Conflict Resolution Program OfficerAlternative Dispute Resolution Mediator (ADR Mediator)
Holding it up: judgment & accountability . Weakest point: embodiment .
Tasks largely resist digitisation Caucusing separately with each side, deciding when a party's stated bottom line is theatre, and calling the moment to push a number are live judgment calls; the automatable slice — issue lists, exhibit indexing, damages spreadsheets, and Modria-style small-claims flows — is real but sits around the edges of a mediation day, which is why this lands at 14 rather than 17.
Some physical or field component An 8 reflects the fact that a hearing room, private breakout spaces, and physical presence at a plant walkthrough or site inspection still matter for many labor and construction disputes, but post-2020 a large share of mediations run entirely on Zoom with e-signed agreements, so the body is useful, not required.
Licensed human required and personally liable Arbitrators issue awards enforceable under the FAA and are named on court, AAA, JAMS, or FMCS rosters with vacatur exposure for evident partiality — but arbitral immunity is broad, most states require no arbitrator license, and many mediators need only a 40-hour training plus court-roster approval, which caps this at 11 instead of the 16+ a bar-licensed practitioner would carry.
Exists to be accountable for ambiguous calls You decide admissibility without rules of evidence, weigh witness credibility with no jury, and issue a final award with essentially no appeal on the merits — an employment or construction arbitration can end someone's career or shift millions on your unreviewable reading of an ambiguous record.
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 (11/20) is whether the law requires a licensed human to sign. Trust premium (17/20) is whether buyers specifically pay for a person. Judgment and accountability (18/20) 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 46 of this occupation's 68 points (68%).
Embodiment (8/20) 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.
Genuine two-tier structure: if small-claims and consumer/e-commerce ODR volume (Amazon, eBay, insurance subrogation, EU ODR-style platforms) is absorbed by automated resolution, the residual caseload is high-stakes multi-party commercial, labor grievance, and family disputes where the work is caucusing, credibility reading, and impasse-breaking — raising the share of the day AI cannot do.
Explicit statutory or rule-level requirement that an arbitral award or mediated settlement be issued and signed by a named natural-person neutral, with AI-generated awards unenforceable. Watch the Revised Uniform Arbitration Act drafting process, state UMA amendments, and court-annexed ADR program rules (e.g., federal district court mediator rosters) adding 'no generative AI shall render the award' certification clauses; also AAA/JAMS roster ethics rules requiring the neutral to certify personal deliberation, mirroring judicial standing orders on AI use.
Vacatur case law: an appellate decision setting aside an award because the neutral delegated reasoning to an AI tool (analogous to 'failure to hear evidence' or evident partiality under FAA s.10) would make personal, documented human deliberation a condition of enforceability.
Party-choice institutionalized: union collective bargaining agreements and commercial arbitration clauses naming an agreed human panel (e.g., FMCS/AAA labor panels, NFL/MLB-style permanent umpires) and expressly excluding algorithmic neutrals. Watch AFL-CIO affiliate contract language and ABA Section of Dispute Resolution model clause updates.
Little headroom at 18; would only rise if neutrals absorb more consequential authority, e.g., expansion of mandatory court-annexed arbitration or med-arb authority where the neutral both mediates and then binds, concentrating the decisive call in one named person.
The limit. Embodiment has no route — the work is a conference room or Zoom and remote ADR is now normalized. The volume risk is not displacement of the elite neutral but collapse of the entry tier: if consumer and low-value disputes go to automated ODR, scores per remaining worker rise while headcount in a 9,210-person occupation shrinks. High score, thin pipeline.
| New York-Newark-Jersey City, NY-NJ | 600 | $77,430 +3% |
| Dallas-Fort Worth-Arlington, TX | 210 | $56,340 -25% |
| San Francisco-Oakland-Fremont, CA | 200 | $108,150 +43% |
| Washington-Arlington-Alexandria, DC-VA-MD-WV | 190 | $109,920 +46% |
| Austin-Round Rock-San Marcos, TX | 170 | $57,160 -24% |
| Philadelphia-Camden-Wilmington, PA-NJ-DE-MD | 140 | $74,520 -1% |
| Phoenix-Mesa-Chandler, AZ | 130 | $82,150 +9% |
| San Jose-Sunnyvale-Santa Clara, CA | 130 | $129,180 +71% |
| San Jose-Sunnyvale-Santa Clara, CA | 130 | $129,180 +71% |
| Albany-Schenectady-Troy, NY | 60 | $112,620 +49% |
| Washington-Arlington-Alexandria, DC-VA-MD-WV | 190 | $109,920 +46% |
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 68. 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.