SAFE
The job is watching water and slopes with your own eyes and then physically entering them — swimming a limp swimmer to the deck, applying rescue breaths, sledding an injured skier down a run, setting avalanche charges. AI vision systems (drowning-detection cameras, avalanche forecasting models) are genuinely encroaching on the surveillance half of the work, but they can only page a human to do the extraction, and health codes plus insurers require certified bodies on duty in fixed ratios. The real pressure on this occupation is budget and seasonality, not software.
Tasks largely resist digitisation. A drowning-detection camera can flag a swimmer on the bottom, but nothing else in the shift — the 20-yard approach swim, the spinal board extraction with a two-person roll, the toboggan belay down a mogul field, the chairlift evacuation rope work — has a digital substitute, which is why this sits at 17 rather than in the mixed band.
Hands-on in uncontrolled environments. You work in moving water, on ice, in whiteouts and 95-degree pool decks, and the environment is the hazard you are managing: 19 reflects that the rescue itself happens in the uncontrolled medium, with the only reason it isn't 20 being the fraction of hours spent scanning from a stand or checking chemical logs.
Certification preferred, not legally required. Lifeguard/CPR/AED and OEC or EMT-B certifications are legally required to occupy the post — health codes set guard-to-bather ratios — but they are short-course certificates renewed every 1-2 years, not a state licence with a personal practice you can lose and be sued over, so the patrol or aquatic operator absorbs the negligence claim, putting this at 8 rather than an RN's 14.
Some relationship component. Patrons don't ask for you by name and rarely learn it, but season-pass skiers and swim-team regulars know the patrol shack faces, and enforcing closures or clearing a slide path depends on riders believing the person in the red jacket has skied that terrain — relationship as compliance tool, not as the product, hence 11.
Meaningful discretion. Deciding to close a run for instability, calling for a helicopter versus a toboggan, or choosing to hold C-spine on a skier who insists he's fine are calls made in seconds with no supervisor present and real death exposure, but they run through protocol trees — OEC algorithms, avalanche hazard scales, EAP triggers — which keeps it at 13 instead of the 16+ of someone setting the protocol.
Has AI actually changed your work?