SAFE
The core of this job is physically managing and teaching 3-to-5-year-olds with autism, speech delays, and motor impairments — hand-over-hand prompting, toileting support, de-escalating meltdowns, running play-based interventions on the floor. AI can draft IEP goal language, summarize progress data, and generate parent communication, but it cannot conduct a developmental observation, hold a distressed child, or sit accountable in an eligibility meeting. State special education licensure plus IDEA's legal requirement that a qualified provider deliver and document services locks a credentialed human into the role.
Dipped in 2020, then grew past where it started.
Median pay $60,000 → $64,830 -13.6% 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
+1.4% 29,300 → 29,700 on the projections basis
Hard to automate, and growing
The work resists current AI and the BLS projects +1.4% 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.
~2,100 openings a year on average, including replacing people who leave.
TeacherDeaf TeacherBlind TeacherBraille TeacherInterventionistAutistic TeacherOutdoor EducatorResource TeacherSpecial EducatorInclusion TeacherElementary TeacherReading SpecialistBehavior SpecialistHandicapped TeacherLearning SpecialistLip Reading TeacherResource SpecialistHome Therapy TeacherEarly InterventionistPublic School TeacherResource Room TeacherSign Language TeacherSpecial Needs TeacherDeaf Education Teacher
Holding it up: embodiment . Weakest point: liability shield .
Tasks largely resist digitisation Hand-over-hand prompting through a fine-motor task, running discrete trial or naturalistic teaching sessions with a nonverbal four-year-old, and taking live data on behavior antecedents while the behavior is happening are the bulk of your contact hours, and none of that survives being moved to a screen — the 15 rather than 19 reflects that IEP paperwork, progress-report narratives, present-levels drafting, and Medicaid billing notes are genuinely being pulled into software.
Hands-on in uncontrolled environments You spend the day on the floor mat, on the toilet-training schedule, physically blocking a child from bolting the classroom, positioning a student in a stander or gait trainer, and doing safe-hold or de-escalation with a child who is biting — an uncontrolled environment where the 'workspace' is a moving three-year-old with sensory dysregulation.
Licensed human required and personally liable State early-childhood special education licensure is a hard prerequisite, and IDEA Part B requires that services be delivered by personnel meeting state qualifications with your signature on the IEP as the provider of record; the 15 rather than 19 is because the district, not you personally, absorbs due-process complaints and compensatory-education orders — you are named, not sued.
Exists to be accountable for ambiguous calls You decide in real time whether a behavior is escape-motivated or pain-driven, whether to abandon a planned activity, when to call a suspected-abuse report, and you sit in eligibility and reevaluation meetings arguing for or against services on ambiguous developmental data for a child too young to have a stable profile — the 16 rather than 19 reflects that the IEP team, not you alone, formally owns the placement decision.
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 (15/20) is whether the law requires a licensed human to sign. Trust premium (18/20) is whether buyers specifically pay for a person. Judgment and accountability (16/20) 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 49 of this occupation's 82 points (60%).
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 91/100, still SAFE.
Task-mix shift: if AI absorbs the documentation tier (progress monitoring narratives, IEP boilerplate, Medicaid billing notes, parent updates) that currently consumes large shares of the week, the residual job is nearly all live observation, behavioral de-escalation, hand-over-hand instruction, and family conferencing — the tier with no automation path. This tier genuinely exists and is already the bottleneck.
State education agencies or OSEP guidance explicitly requiring that any AI-generated IEP goal, progress note, or eligibility determination be reviewed and personally signed by the licensed special educator of record, with the signature carrying due-process liability — mirroring the district-level AI policies already being adopted (e.g., state DOE AI guidance in CA, OR, WA) and the OCR/OSEP position that IDEA obligations rest on a qualified provider. Also plausible: a Due Process hearing decision voiding an IEP written substantially by AI without documented educator review.
If eligibility and least-restrictive-environment placement calls become more contested — rising due-process filings and mediation rates plus mandated documentation of the educator's independent clinical rationale for departing from an AI/data-system recommendation — the role formally owns the ambiguous call rather than ratifying a system output.
Little headroom; already near ceiling. A minor rise only if staffing rules raise the ratio of one-on-one physical intervention time (e.g., state mandates on restraint/seclusion requiring two credentialed adults present during physical management, as in several state restraint statutes).
Parents of children with disabilities already select programs on the named teacher and advocate for human-delivered service hours in IEP negotiation; a formal parental right to refuse AI-mediated instruction or AI-generated assessment inputs, written into state IDEA implementation rules, would harden that preference into a purchasing constraint.
The limit. Already 82/100 with embodiment and trust near maximum; realistic total headroom is roughly 8-9 points. The binding risk to this occupation is not AI but funding and enrollment — IDEA Part B preschool appropriations, district budget cuts, and paraprofessional substitution — which the register does not score.
| New York-Newark-Jersey City, NY-NJ | 6,110 | $81,210 +25% |
| Chicago-Naperville-Elgin, IL-IN | 900 | $61,950 -4% |
| Minneapolis-St. Paul-Bloomington, MN-WI | 760 | $80,700 +24% |
| Dallas-Fort Worth-Arlington, TX | 750 | $74,110 +14% |
| Boston-Cambridge-Newton, MA-NH | 610 | $99,440 +53% |
| Philadelphia-Camden-Wilmington, PA-NJ-DE-MD | 510 | $79,560 +23% |
| Seattle-Tacoma-Bellevue, WA | 470 | $94,240 +45% |
| Washington-Arlington-Alexandria, DC-VA-MD-WV | 440 | $82,760 +28% |
| Kansas City, MO-KS | 380 | $119,560 +84% |
| San Diego-Chula Vista-Carlsbad, CA | 240 | $117,200 +81% |
| Akron, OH | 60 | $100,650 +55% |
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 82. 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.