EXPOSED
Lecture writing, slide decks, syllabus construction, quiz banks, reading summaries and first-pass essay feedback are all things current models do at usable quality, and much of an intro human/physical geography course is exactly that. What persists is live classroom facilitation, GIS and remote-sensing lab supervision, field methods instruction, thesis advising and the institutional judgment of grading and curriculum design. The bigger near-term threat to this 3,300-person occupation is departmental consolidation and enrollment decline in small geography programs, which AI-assisted course delivery accelerates rather than causes.
Mixed — a routine tier and a judgment tier. Roughly half the workload — GIS lab troubleshooting when a projection fails mid-exercise, running a field methods week with GPS units and soil augers, defending a thesis proposal with a committee, reading a student's confused mental map and diagnosing why — has no text-output substitute, while the intro-course scaffolding of lecture notes, quiz banks and map-reading worksheets already does, which is what pins this at 11 rather than up with lab scientists or down with content writers.
Some physical or field component. Field camp, campus watershed walks, hauling total stations and Trimble units outdoors, and standing over a student's shoulder at an ArcGIS workstation are real physical duties, but they are a few weeks and a few lab hours per term against a base of lecturing, screen-based mapping and committee work — a 9, not the 15 of a field geologist who lives in the terrain.
No licence, no signature requirement. No state licence, no board exam, no professional registry gates who teaches human geography — a PhD and a departmental hiring vote are the only entry controls, and GISP certification is optional and unenforced even for the GIS courses, so nothing in law requires a credentialed human to deliver the content.
The human relationship is the product. At 14 because the durable value is the named relationship: students choose a thesis advisor, ask for the letter of recommendation that gets them into a planning MA or a federal GIS post, and return years later for the reference — the lecture content is fungible, the advisor's signature and knowledge of that student's work is not.
Meaningful discretion. Grading a contested essay on political borders, setting curriculum for an accreditation review, deciding whether a student's fieldwork is publishable, and calling safety judgments on a field trip are genuine discretion, but they sit inside departmental grading policies, catalogue requirements and IRB review rather than being unappealable high-stakes calls — hence 10, mid-band.
Has AI actually changed your work?