EXPOSED
The bulk of an econ professor's week — building lecture decks, writing problem sets and exams, grading, summarizing literature, running regressions and drafting papers — is exactly the text-and-numbers work current models do at usable quality, and online sections are the most exposed of all. What persists is the live seminar, thesis and dissertation supervision, letters of recommendation, and being the named human who assigns a grade and defends a research claim. The real threat here is structural: enrollment pressure, adjunctification, and larger sections shrink headcount faster than AI replaces the teaching act itself.
Mixed — a routine tier and a judgment tier. Micro/macro problem sets, multiple-choice midterms, Stata/R regression scripts and lit reviews are all reproducible by a model today, which is why this sits at 9 rather than mid-teens; the residue that holds — running a live 15-person seminar where you push back on a student's identification strategy, reading a dissertation chapter that isn't working yet, and writing the recommendation letter a hiring committee will actually weigh — is real but is a minority of contact hours in a 200-seat Principles course.
Some physical or field component. The physical demand is standing in front of a lecture hall for three 50-minute sessions, working a chalkboard or clicker, and holding office hours — a controlled indoor environment with no equipment to handle, which is why this is closer to the desk end than to lab-science or clinical faculty who must be in a specific room with specific apparatus.
No licence, no signature requirement. There is no state licence to teach economics; the credential is a PhD and a departmental hiring vote, and accreditors (SACSCOC, AACSB for b-schools) audit faculty qualifications at the program level, not by issuing you a personal ticket that can be pulled — so nothing legally requires a named human to deliver the content of ECON 101.
The human relationship is the product. At 14 the relationship is doing the work that the content cannot: students choose sections by professor, dissertation advising is a multi-year named commitment, your recommendation letter is trusted because a specific economist with a reputation signed it, and consulting or expert-witness work follows the person out of the department — none of that transfers to whoever inherits the syllabus.
Meaningful discretion. Grade curves, academic-integrity referrals, deciding whether a dissertation is defensible, and choosing what to claim from noisy identification are genuine discretion, but they run inside faculty handbooks, departmental grade norms and IRB/committee review, and a wrong call is appealed rather than catastrophic — that combination lands at 10, not at the level of someone whose single judgment ends a career or a life.
Has AI actually changed your work?