EXPOSED
Law professors do two things AI is already good at — producing doctrinal explanation and drafting written analysis — and one thing it is not: running a live Socratic classroom where students are pushed to defend positions under pressure, and then vouching for those students to employers and courts. Case summaries, model answers, exam-question banks, outline generation, and first drafts of law review pieces are all substantially automatable today; cold-calling, clinic supervision, moot court coaching, and letters of recommendation are not. The real protection here is institutional rather than technological: ABA accreditation standards require full-time faculty, tenure limits headcount churn, and the bar-passage pipeline is a credentialing monopoly — all regulatory scaffolding that can be revised.
Mixed — a routine tier and a judgment tier. Doctrinal lecture prep, casebook note-writing, hypothetical drafting, and grading multiple-choice or IRAC-formatted answers are all reproducible today, which pulls this below the safe band, but cold-calling a 1L through a line of argument he hasn't thought through, supervising a live clinic docket with real clients, and coaching an oral argument keep it at 12 rather than 6.
Some physical or field component. The work is classroom, office hours, and Zoom, so 7 reflects only the physical fragments that do exist — being bodily present in a lecture hall for accreditation-relevant contact hours, supervising a clinic that appears in court, and judging moot court rounds in person — none of which involves uncontrolled environments.
Certification preferred, not legally required. Most law professors hold a JD and many a bar licence, but they are not personally liable for teaching content and no statute requires a licence to lecture on torts; the 8 comes from clinical faculty, who are the actual attorney of record on client matters and answer to state bar discipline for them.
The human relationship is the product. Students choose seminars by professor, judges hire clerks on a specific professor's phone call, and a recommendation letter's value is entirely who signed it — the reputational vouching function cannot be transferred to another instructor or a system, which is why this sits at the top of the band.
Exists to be accountable for ambiguous calls. Deciding whether a student's professional-responsibility lapse in clinic gets remediated or reported, setting the curve that determines who makes law review, and choosing what unsettled doctrine to teach as the rule are calls with no procedure behind them and consequences that follow the student into practice.
Has AI actually changed your work?