■ SAFE RISK ■ Public Service & Government
No. Machines are getting very good at finding the casualty and very bad at getting them off the mountain, and the second half is the job.
“Drones locate hikers. But carrying someone down a scree slope in a blizzard? That's you.”
Our AI replacement risk score — how we score jobs
A callout starts with a phone alert and a grid reference that is usually wrong. The team leader builds a search plan from weather, last-known-position, the caller's description of the party's kit, and a hard-won sense of where lost people actually go — downhill, toward water, into the wrong gully. Then twelve to thirty volunteers walk hillsides in the dark, carrying stretchers, vacuum splints, ropes and hot drinks, often for six hours before anyone is found.
The technology has genuinely moved. Thermal drones sweep a corrie in minutes that would take a hasty team an hour. Phone-location handovers from the emergency services, what3words fixes, avalanche transceiver searches and mapping software with automatic probability-of-area modelling have all compressed the search phase. Some teams now run drone pilots as a specialist role alongside dog handlers and swiftwater technicians. Search — the part that is fundamentally a sensing and inference problem — is where automation is winning, and rescuers welcome it, because searching is the boring part.
Everything after the find resists hard. Packaging a hypothermic casualty with a suspected pelvic fracture onto a stretcher, rigging a lower down a wet slab, managing a rope system while wind gusts rip at the anchor, and carrying two hundred kilos of person-plus-kit across boulder fields is physical work in terrain built to defeat wheels. Casualty care is judgement under cold, dark and time pressure with incomplete information. And the whole structure is volunteer and social: local knowledge, team cohesion, standing in the rain at 3 a.m. because someone had to. Nobody is funding a robot for that, and if they did, it would still need someone to carry it up.
Automatability: our editorial assessment of current and near-term AI capability
Expect the search phase to keep shrinking through the 2030s as drone fleets and location handovers improve, which means fewer wasted hours and possibly smaller callouts. The evacuation, medical and leadership work stays human well past 2040 — the terrain, the weight and the unpredictability of injured people are exactly what robotics has failed at for thirty years. Our risk score of 17 reflects tooling change, not job loss.
No, they change what volunteers do. Drones compress the search phase, which is where most of the hours used to go, but every callout still ends with humans stabilising a casualty and physically moving them off steep ground. In practice teams that adopt drones do not shrink — they reallocate people to medical, rope and control roles.
Yes. Demand tracks hillwalking participation, which keeps rising, and the skill stack — rope systems, pre-hospital care, navigation, team leadership — transfers directly into paid outdoor instruction, expedition medicine and emergency services work. Nothing on the horizon replaces the carry-out. Our risk score of 17 is among the lowest we assign.
Thermal-capable drones with a trained pilot roster, reliable phone-location handover with the police, and mapping software that logs search coverage so you can prove what ground has been cleared. Those three cut search time substantially. Spend the savings on medical training and stretcher-handling practice, which is where callouts are still won or lost.
Not usefully, and not soon. Legged robots still struggle on loose scree and wet rock while carrying a fraction of an adult's weight, and a casualty with a spinal or pelvic injury needs continuous handling adjustments no current system can make. Helicopters already solve the cases where machines can help; the rest is why teams exist.