■ SAFE RISK ■ Public Service & Government
No. Robots are getting very good at finding people and delivering flotation; nobody has built a machine that can wrestle a panicking, hypothermic adult out of breaking surf.
“Drones drop rafts. But swimming into raging water to grab someone? Insanely human.”
Our AI replacement risk score — how we score jobs
The role is high-consequence and mostly preparation. Endless conditioning, pool and open-water drills, helicopter deployment training, equipment checks on masks, fins, harnesses and radios, and medical currency in resuscitation and hypothermia management. On a live call it becomes a rapid sequence: assess sea state and hazards, deploy from a hovering aircraft or vessel, approach a person who may fight you, apply a rescue hold, secure them into a strop or basket, and manage the casualty's airway and temperature during recovery — often while the aircraft is running low on fuel and the weather is deteriorating.
Automation has already changed the search half. Drones cover coastline and open water far faster than a helicopter sweep, with thermal imaging and computer vision that spots a head in water better than a tired human eye. Remotely operated flotation devices — motorized buoys that can be driven to a swimmer in seconds — are in service on beaches and demonstrably save lives. Drift-prediction models narrow search areas dramatically, and autonomous surface vessels can hold station in conditions that would be dangerous to crew. Every one of these reduces the number of dives a human has to make, which is exactly the point.
But delivering flotation is not the same as extracting a person. Real casualties are unconscious, injured, trapped in rigging, tangled in a capsized hull, or so panicked they will drown their rescuer. Handling that requires strength, in-water judgment, and the ability to make a physical decision under load — including the decision to let go. Add casualty medical care, multi-casualty triage in the water, and the confined-space work of entering a flooded vessel, and you have a task set well outside current robotics. The role will keep shrinking in frequency as remote tools catch the easy cases, while the calls that remain get harder. Our score reflects a job that is being made rarer and more specialized, not obsolete.
Automatability: our editorial assessment of current and near-term AI capability
The search and flotation-delivery layers are automating right now — drones and remote rescue buoys are already standard on many coastlines and will be near-universal by 2030. Human swimmers will deploy less often but on harder jobs. Through 2040 the role narrows toward complex extraction, casualty care and confined-water work, which remains outside robotic reach.
They are replacing some deployments, which is a good thing. A drone that drops a flotation device to a conscious swimmer in ninety seconds beats putting a human in the water. What drones cannot do is grab an unconscious casualty, free someone tangled in rigging, or manage an airway. The result is fewer, harder swims per swimmer.
Yes, though the shape of it is changing. Demand for coastal and maritime rescue is not declining, and crews are adding remote-asset roles rather than shedding swimmers. Expect job descriptions to include drone piloting and sensor interpretation alongside water competency. The physical standards and the medical training remain the entry barrier.
Search large areas with thermal and visual detection, deliver motorized flotation buoys to swimmers within seconds, hold station as a surface platform in rough conditions, and narrow search zones with drift modelling. Those are meaningful, life-saving contributions. Extraction, medical care and any casualty who is unconscious or resisting still require a person.
Advanced pre-hospital medicine, technical rescue in confined and swiftwater environments, and remote-asset operation. The swimmer who can also run the drone, read the drift model and provide casualty care is the one the crew builds around. Pure fitness remains necessary but is no longer the differentiator it once was.