Why this work stays in human hands
Engine and other machine assemblers build complete machines from loose parts. The day is spent positioning blocks and housings, fitting pistons and gears, bolting and torquing fasteners, aligning shafts, then running the finished unit to listen for noise and check for leaks. Those are physical judgments made in tight spaces with heavy, oily parts.
Software is good at the paperwork around that work. It can read a build sheet, flag a torque reading that drifts, or sort inspection photos. It cannot feel a bolt start to bind, notice that a gasket sits proud by a hair, or decide that a part should be reworked rather than shipped. That gap is why the question of whether AI will replace machine assemblers has a different answer than it does for desk work.
Scale matters too. This is a small occupation: about 34,000 jobs in the United States, with median pay of $53,710 a year (BLS, 2025). The same source projects employment down 17% between 2025 and 2035 (BLS, 2025). Fewer openings is a real pressure. It is not the same thing as the tasks moving to a machine.
What AI does, what it helps with, and what people keep
Our coverage figure, which answers “Can AI do it?”, sits at 7 on a 0 to 100 scale where higher means more task time AI can handle today. You can read how that is built on the coverage method page.
The slice AI can take on its own is 0%. It is clerical edging: logging build and test results, keeping part counts and traceability records straight, and summarizing defect reports for a supervisor. None of it touches the engine.
The assist slice is 10%. Here a human still does the job, with software alongside: pulling up the right specification or exploded view while reading blueprints, checking a measured dimension against tolerance, or watching sensor data during a test run so a fault gets caught earlier.
The rest, 90%, is the job as most people picture it. Lifting and seating components with a hoist, fastening and aligning them by hand, verifying fit with gauges, and reworking an assembly that did not pass. Hands, eyes and a feel for the machine do that part.
What has actually been tested
No study has measured an AI system against people doing this job end to end. Our quality-parity grade is D, which is the grade we use when there is no direct head-to-head test, so we publish no parity number for engine assembly. Saying otherwise would be guessing.
What would settle it is specific: a timed build of the same engine or machine family, same tolerances, same first-pass yield and rework rate, by a robotic cell and by qualified assemblers, with the results published. Vision-guided assembly demos exist in the industry, but a demo on a fixed part is not a measured comparison across mixed models. Until that exists, the honest position is an open question, and our quality-parity method explains why a grade like this never gets a score.
When the picture could change
Most likely after 2046 (8 in 10 of our scenarios). The replacement-year method sets out what that window does and does not claim.
Two things could pull it earlier. First, consolidation: if engine building concentrates into a few high-volume plants, the economics of fixed automation improve, and the automation already used on this work is of that fixed kind rather than flexible. Second, cheaper machine vision and force sensing, which would let a cell handle parts that vary slightly instead of only identical ones.
Two things hold it back. Capital and changeover are the first. Running software is cheap next to a wage; fixtures, cells and reprogramming for a new model are not, and this is low-volume, high-variety work in many shops. The second is the share of the job that is physical, which is nearly all of it. That is a hardware problem more than a model problem, and our guide to robots and physical jobs walks through how slowly that hardware has moved.
Good to know: this job sits inside a wider manufacturing trend, so read the headline number together with the employment projection rather than on its own.
How to stay needed
The safest ground is the work in the human slice above. Lean into three parts of it. Test and diagnosis: being the person who can run a machine, hear what is wrong and say why. Rework: taking an assembly apart, finding the cause and putting it right. Precision fitting and alignment on parts that vary, where a gauge reading has to be judged, not just recorded.
Two skills travel well from there. One is reading and interpreting technical drawings and specifications well enough to spot an error in the paperwork. The other is working around automation: loading, tending and troubleshooting an automated cell, including the basics of why it stopped. Both make you the person a plant keeps when a line changes.
Nearby jobs worth a look are electromechanical equipment assemblers, electrical and electronic equipment assemblers and team assemblers. The whole group sits on the assemblers and fabricators family page, and the wider manufacturing sector page shows how the rest of the plant scores.
Our headline figure here, Still needs a human, is 85 out of 100 (higher is safer). How we reach it is set out in our methodology. From there you can put this job side by side with another, or see where it lands among the jobs that mostly need a person.