A boiler does not.
Neither does a roof leak, a bad circuit, a cracked weld, a failed compressor, a clogged drain, or an elevator stuck between floors.
The physical world has a way of resisting abstraction. Old houses are not standardized. Crawl spaces are cramped. Wires are mislabeled. Pipes were repaired by someone else 30 years ago. A machine can throw a fault code and still miss the failure.
That is the trade worker’s advantage. The job has to be done somewhere, on something, with consequences.
But that does not make the trades immune.
It only changes where the machine enters.
AI may not replace the electrician at the panel, but it can change the work before he gets there and after he leaves: the estimate, the dispatch, the permit check, the diagnostic report, the safety checklist, and the invoice. It can read plans, schedule crews, flag delays, order parts, and monitor a jobsite through cameras, sensors, drones, and digital models.³
The worker still shows up.
But the work around the worker changes.
Joseph Paredes, a 26-year-old Texas welder profiled by Reuters, already works in a shop where robots are part of the picture. That did not make him see the trade as obsolete. It made him see the boundary more clearly. “We still need a physical person, a human, to make those robots work,” he said.⁴
That is probably the better way to understand the next decade. AI will not move through the trades the way it is beginning to move through office work. It will arrive first where trade work becomes standardized, repeatable, measurable, and software-visible.
Factory-built housing. Automated welding. Drone inspection. Predictive maintenance. Smart buildings that report their own failures.
It will arrive more slowly where the work is old, wet, cramped, dangerous, improvised, or legally accountable.
That means the safest trades may not be the least technical. They may be the most technical.
Electricians, HVAC technicians, elevator mechanics, industrial maintenance workers, grid workers, and controls specialists are not protected because their work is old-fashioned. They may be protected because their work sits at the intersection of the physical and the digital. The job requires hands, but also judgment. It requires tools, but also software. It requires showing up, but also knowing what the machine is telling you — and when the machine is wrong.
This is also why the AI economy may increase demand for some trades before it reduces it. Data centers, chip plants, battery systems, power lines, substations, heat pumps, solar arrays, and grid upgrades do not build themselves. Google has backed an effort to train 100,000 electrical workers and 30,000 new apprentices in the United States, not as charity, but because the cloud still needs electricians.⁵
The invisible economy has a very visible footprint.