EXPOSED
Roughly half the job is screen work AI already does competently — literature reviews, species distribution modeling in R, population statistics, permit paperwork, grant boilerplate, and the endless technical reports and NEPA/ESA documentation. The other half is genuinely embodied: mist-netting, radio-collaring, scat and pellet transects, necropsies, boat and backcountry surveys in weather that doesn't cooperate. There's no license gate — 'wildlife biologist' is a job description, not a protected title — and camera traps plus automated bioacoustic and image classifiers are already shrinking the labor needed per survey.
Mixed — a routine tier and a judgment tier. Camera-trap image sorting, BirdNET-style acoustic ID, occupancy models in unmarked, and the first draft of a biological assessment are all things software now does at speed, but nobody has automated setting a mist net at the right height in the right flyway, sexing a sedated bear, or reading a fresh browse line — the 11 reflects that the field half stays and the analysis-and-writing half is already leaking.
Hands-on in uncontrolled environments. A 14 covers work where the sampling site is the office: hip waders in spring runoff, elk capture from a helicopter, scat transects on unmaintained trails, necropsy on a carcass in whatever condition you found it — uncontrolled ground and weather, though not the sustained physical extremity of a smokejumper or commercial diver.
No licence, no signature requirement. There is no state licensure for zoologists; the 4 exists only because collection, banding, and Section 10 take permits are issued to a named individual with demonstrated competency, so USFWS or a state agency can pull your authorization — but the permit follows the project, not a protected title, and a technician or contractor can hold it.
Some relationship component. A 7 reflects that landowners granting survey access, tribal co-management partners, and the agency biologist who has to defend your numbers at a public comment hearing all deal with you specifically, but the deliverable is a dataset and a report that another qualified biologist could have produced identically.
Meaningful discretion. The 11 sits on real calls — declaring a nest active, setting a take threshold, deciding a population estimate is robust enough to underwrite a listing or delisting decision — but those calls go through agency review, peer review, and established survey protocols before anyone acts on them, so you are rarely the last signature.
Has AI actually changed your work?