Why the work stays in the water with people
Commercial diving is repair work done in a place that punishes mistakes. Divers inspect and test submerged structures, weld and cut metal at depth, rig and move heavy loads, and patch pipelines, pilings and hulls. Before any of that, someone decides whether the water, the current and the structure allow a dive at all. That judgment is made on the boat, by people who will live with the result.
The second reason is the hands. A diver works by feel in low visibility, threading a bolt, clearing marine growth, setting a clamp against swell. Our robotics read for this job puts most of the task time in the physical column and rates the capability needed at the dexterous humanoid tier, which is the hardest class of machine to build and the furthest from routine field use. Remotely operated vehicles are real and widely used, but they are mostly eyes, lights and a manipulator arm under human control, not a self-directed repair crew.
So the honest answer to the question people type, will AI replace commercial divers, is that software is reshaping the paperwork and the survey end of the trade while the wet work stays human. The share of task time that still needs a person here is 94%, and the tasks in that group are the ones the industry pays for: going in, fixing something, and coming back.
What AI does, what it assists, and what it leaves to divers
The tasks our data puts in the AI-does group are desk-side: writing up inspection records and sorting or tagging survey footage so an engineer can find the frame that matters. That slice of task time is 0%. It is the part of a diver’s day that happens dry, after the dive.
Assistance is broader in effect than in size. Software helps plan dives and gas schedules, flags corrosion and cracks in video from a camera or vehicle, and keeps logs tied to position data. The assisted share of task time is 6%. Machine vision can mark a weld seam as suspect; a diver or a supervisor still decides whether to cut it out. If you want the detail behind this split, our coverage score method explains how task time is counted.
What is left to people is the trade itself: underwater welding and cutting, equipment checks and gas management, rigging and salvage, and communication with the surface while conditions change. Confined-space entry, zero-visibility work by touch, and emergency response all sit here too. Our overall coverage figure for this job is 7 out of 100, which reflects how little of the work a system can take end to end today.
What the evidence covers, and what it does not
No study has put an AI system against a working commercial diver and measured who did the job better. The evidence grade here is D, our lowest tier, which means the quality comparison has not been measured. We do not publish a parity number when that is the case, and you should treat any site that does with suspicion.
What would settle it is specific and testable: a controlled trial of autonomous inspection against diver inspection on the same structures, with defect-detection rates reported; a demonstration of unmanned underwater welding or clamp installation to code, without a diver standing by; and incident data comparing machine-only and diver-assisted operations. Until something like that exists, the fair statement is that machines have taken survey and observation hours, not repair hours. The wider method, including how we grade evidence, is on our scoring methodology page.
The labor data is steadier ground. The Bureau of Labor Statistics counts about 3,450 commercial divers in the United States, with median pay of $72,990 (BLS, 2025) and projected employment growth of 4.8% from 2025 to 2035. That is a small, specialized trade growing slowly, not a shrinking one.
When subsea work could shift to machines
Most likely after 2046 (8 in 10 of our scenarios). Our replacement-year method explains what that window measures and how the spread is built.
Two things could pull it earlier. Cost is the clearest: the cost panel above compares machine time with a crewed dive spread, and the gap gives operators a strong reason to send a vehicle wherever one will do. Safety rules point the same way, because every hour a person is not in the water at depth is an hour of risk removed.
Two things hold it back. Dexterity is the first: resident vehicles and arms can brush, scan and film, but setting, torquing and welding against current is still a human skill, and our robotics tier reflects that. The second is the economics of a small market. Fewer than four thousand divers nationwide (BLS, 2025) is a thin commercial prize for anyone building a purpose-made subsea repair robot, so investment follows oil and gas inspection instead.
Good to know: ROV growth has been visible in this trade for decades, and the diving workforce has kept growing alongside it.
How to stay needed in this trade
Lean into the work that keeps you on the critical path. Three areas stand out: underwater welding and cutting to code, rigging and salvage on live structures, and dive supervision, including gas planning, equipment checks and emergency response. Those are the tasks our list keeps firmly in the human column.
Two skills raise your value quickly. The first is running and interpreting the machines: pilot a remotely operated vehicle, read sonar and video, and know when the footage is lying. The second is inspection reporting, including non-destructive testing methods, so you can sign off findings an engineer will act on. Divers who can do both the dive and the data become the person a client calls first.
If you are weighing nearby trades, the closest work on deck and in marine operations is worth a look: riggers, ship engineers and sailors and marine oilers. You can also see where this job sits among other maintenance and repair occupations, read the picture for the oil and gas sector, browse the jobs that mostly need a person (our top band, Nah.) on our safest jobs list, or put two trades side by side with the job comparison tool.