SAFE
The paperwork half of this job — drafting IEP goal language, progress-monitoring summaries, accommodation checklists, parent update letters — is squarely in AI's wheelhouse and will get faster. The other half is bodily and unscheduled: de-escalating a meltdown in a hallway, prompting a 12-year-old through a math task hand-over-hand, reading a nonverbal student's shutdown before it becomes a crisis, and holding a contentious IEP meeting together. State licensure plus IDEA's legal machinery require a named, certified human to author and sign the IEP and to be answerable in due-process hearings.
Tasks largely resist digitisation. The generative gains land on goal-bank drafting and progress notes, but the daily core — running small-group reading in a resource room, co-teaching a 28-student general-ed math block, implementing a behavior intervention plan minute-by-minute, and taking frequency data on a student who elopes — is bodily supervision of adolescents that no system performs, which puts it at 14 rather than the 17+ of trades with no paperwork at all.
Hands-on in uncontrolled environments. You are on your feet in hallways, cafeterias, bus lines, and toileting/hygiene routines for students with significant needs, doing physical prompting and sometimes state-certified restraint or escort holds in a building whose layout and crowd you don't control — held below the high teens only because a real share of the week is spent at a desk on IEP documents and meetings.
Licensed human required and personally liable. A state special-education credential is mandatory and IDEA names you as the IEP team member whose signature and data are examined in a due-process hearing or OCR complaint, so a compliance failure — missed timeline, unimplemented accommodation — attaches to you personally; it sits at 14 not 18 because the district and LEA representative carry the financial exposure and defend the case.
The human relationship is the product. Parents who arrived at middle school after years of fighting elementary placements decide whether to consent to services based on whether they trust you specifically, and the student who will attempt a reading task for you and no one else is the intervention — that continuity of one known adult across three years is what the job delivers.
Exists to be accountable for ambiguous calls. You decide in the moment whether a behavior is escape-motivated or a genuine skill deficit, whether to pull a student from inclusion, whether a 13-year-old's refusal is defiance or a manifestation of disability under a manifestation determination review, and whether present-levels data justify exiting a service — calls made with incomplete information that alter a child's placement and are litigated later.
Has AI actually changed your work?