Why aircraft maintenance stays hands-on
Will AI replace aircraft mechanics? Not on what the evidence shows today. The work is built around physical contact with one specific airplane: opening panels, inspecting structure and engines, chasing an intermittent fault through wiring and lines, then torquing parts back to spec and checking the result. Software can read the manual alongside you. It cannot reach into a wheel well, feel a loose fastener, or hear a bearing that sounds wrong on a run-up.
Accountability also sits with a person. A certificated mechanic signs the maintenance record and the return to service. That signature carries legal weight, and it rests on what the signer saw and did, not on what a model predicted. Any tool that writes part of the paperwork still needs a named human to check the aircraft and accept the work.
There is scale behind the job as well. The US employed about 138,090 aircraft mechanics and service technicians, with median pay of $79,870 a year (BLS, May 2024 data, published 2025). BLS projects employment growth of 5.4% between 2025 and 2035. That is steady demand for people who can turn a wrench and judge a crack.
What AI does, what it helps with, and what stays with you
The share of task time AI can handle on its own is 0%. The routine candidates are text and data work around the hangar: drafting entries for maintenance records and work cards, and pulling the right service bulletin or manual section out of thousands of pages. Both are checkable, repeatable and already partly digital.
Assisted work is a bigger slice: 15% of task time. Here, fault diagnosis is the clearest example. Engine and airframe sensor data feeds predictive maintenance tools that flag a trend before a part fails, and drones or camera rigs can photograph skin and structure for review. A model can rank likely causes. The mechanic still opens the panel, confirms the cause and decides what to replace.
Work that needs a person comes to 85% of task time. That means removing and installing engines and components, repairing sheet metal and structure, rigging control surfaces, running engines and troubleshooting what the data did not catch, and signing the aircraft back into service. Most of that task time is physical, and the robotics tier for this job is dexterous humanoid: hardware that would have to work in cramped, awkward, non-standard spaces. That hardware is not in service in hangars. How we measure the automatable share is set out in how coverage is scored.
What the evidence actually tests
Our parity grade for this job is D, which means no study has yet tested an AI system against a qualified mechanic on this job’s real tasks. So there is no parity number here, and we will not give one. The evidence list on this page shows what exists, and the grade changes if a direct test appears.
What would settle it is specific: a trial where a system diagnoses and clears real squawks on real aircraft, under a maintenance organization’s procedures, measured against licensed technicians on accuracy, rework and time. Until something like that is published and repeatable, claims about aircraft maintenance going automatic are projections, not results. You can read how we grade evidence on the quality parity page, and see the full method at how the scoring works.
When the picture could change
Most likely after 2045 (8 in 10 of our scenarios). What that window means, and how we build it, is explained on the replacement-year page.
Two things could pull it earlier. First, sensor coverage: as more engines and systems stream condition data, diagnostic hours shrink and planning shifts to scheduled part swaps. Second, dexterous robotics: if humanoid or mobile manipulators become reliable and affordable enough for repeatable tasks such as panel removal, cleaning and fastener work, some hangar hours move off people.
Two things hold it back. Regulation is the strongest: tools, parts and procedures need approval, and the sign-off stays with a certificated person. Capital cost is the second. The cost panel on this page compares what an AI system costs per task against the human cost, and the gap narrows only where the work is software, not where a machine has to climb into an airframe. Fleet variety, aging aircraft and one-off damage repairs add more friction.
What to do: get fluent with the predictive maintenance dashboards your employer already runs, so you are the one interpreting the alerts rather than only receiving them.
How to stay needed in the hangar
Lean into the parts of the job that stay with people. Troubleshooting intermittent faults that the data does not explain. Structural and sheet metal repair, where the fix depends on what you find when the panel is off. Airworthiness judgment and the sign-off that follows it, including the call to ground an aircraft.
Two skills raise your value fast. One is data literacy: reading trend monitoring output and knowing when a flag is noise. The other is clear technical writing, because a well-written defect report and repair record is what engineering, inspectors and the next shift rely on.
If you want to see how close work compares, look at avionics technicians, aviation inspectors and aerospace engineering technicians. You can also put two jobs side by side on the job comparison tool, browse the wider vehicle and mobile equipment repair family, check the transportation and warehousing sector, or see where hands-on trades sit in the list of jobs that most need a person.
The headline figure for this job is 83 out of 100 (higher is safer). It is a measure of how much of the work still needs a person, not a promise about any one employer or any one hangar.