EXPOSED
A large slice of the job — literature reviews, ration formulation runs, statistical analysis of trial data, grant and journal writing, extension bulletins — is exactly the text-and-tabular work AI now does at usable quality. What holds is the physical half: designing and running feeding, breeding, and reproduction trials in barns and feedlots, collecting samples, handling animals, and owning the call on whether a genetics or nutrition program goes to producers. No license protects the title, so the moat is embodiment and accountability for real herds, not regulation.
Mixed — a routine tier and a judgment tier. Ration balancing against NRC requirement tables, ANOVA/mixed-model runs on trial data, and drafting extension bulletins are already handled well by software plus an LLM, but designing a 90-day feeding trial with valid pen replication, walking the barn to spot a lame heifer skewing intake data, and collecting blood, semen, or rumen fluid still require a person on site — hence mid-band 11 rather than the 14+ of work that cannot be moved to a screen.
Some physical or field component. You are in feedlots, farrowing houses, and research dairies with restraint chutes, AI guns, ultrasound wands, and manure underfoot, but the sample then goes to a controlled lab bench and much of the year is spent on analysis and writing — a split that lands at 12, above a screen-only analyst and below the veterinarian or herdsman whose whole day is hands on live animals.
No licence, no signature requirement. There is no license to practice animal science: a PhD and ARPAS or ANSI board certification are resume assets, not statutory gates, and when a nutrition program goes to producers it is the feed company, the veterinarian of record, or the IACUC that carries the legal exposure — 3 reflects that IACUC protocol authorship is the only place your name binds anything.
Some relationship component. Producers, feed-company clients, and breed associations do keep coming back to the specific scientist whose ration or sire-selection advice worked on their operation, but the deliverable is a peer-reviewed paper, a spec sheet, or a trial report that stands on its numbers regardless of who signed it, so 8 sits above anonymous output and well below the relationship-is-the-product tier.
Meaningful discretion. Deciding that a genomic selection index or a transition-cow diet is ready to release to commercial herds — where a wrong call costs thousands of animals and a client's season — is a genuinely ambiguous high-stakes judgment, but much of it is bounded by NRC requirement tables, established statistical protocols, and IACUC review, which caps this at 12 rather than the 15+ of someone with no external checkpoint.
Dietitians and Nutritionists EXPOSED
Has AI actually changed your work?