Why this work stays at the bench
Medical appliance technicians build one-off devices for one body. A technician takes a cast or scan of a residual limb, shapes a socket, bends and welds metal uprights for a brace, grinds and polishes the plastic, then fits the finished device and adjusts it after the patient stands up and walks. Every step reacts to a person who is right there in the room.
That is the honest reason the score lands where it does. Software can hold the prescription, nest the parts and drive a printer. It cannot feel a pressure point through a check socket, read a wince, or decide that a trim line needs another two millimeters off the medial side. The work is custom, physical and clinical at the same time, which is an unusual combination.
The occupation is small. About 11,280 people worked as medical appliance technicians in the United States, with median pay of $48,030, and employment is projected to change by roughly 3.5% between 2025 and 2035 (BLS, 2025). Small, skilled occupations like this one rarely attract purpose-built automation, because the hardware cost has to be spread across very few benches.
What AI handles, what it assists, what stays with people
The share of task time AI can handle on its own prints above as 0%. The work that sits closest to that line is the paperwork and the geometry: reading and logging work orders, keeping device records, and turning a scan into a printable file. Pattern work and material nesting have been software jobs for years in digital labs.
Assisted work is the larger story. 13% of task time is work where a tool speeds a technician up rather than standing in for one. Digital scanning replaces plaster casting in some clinics, and CAD carving or printing can rough out a socket or an ankle-foot orthosis shell. The technician still selects materials, checks tolerances and signs off on what leaves the lab. If you want the definition behind that split, it is set out in how coverage is measured.
What is left is the bulk of the job: 87% of task time still needs a person. Taking measurements and impressions from a patient, fitting and adjusting the device on the body, and repairing or modifying appliances that come back worn or outgrown all happen with hands on the work. So does the laminating, grinding and heat-forming that turns a shell into a wearable device.
What has actually been tested
No one has run a published head-to-head trial of an AI system against a qualified medical appliance technician. That is why the quality-parity grade on this page prints as D, and why no parity number is given. A grade at that level means the evidence is missing, not that AI quietly won or lost.
A fair test would be specific. Take a batch of prescriptions for the same device type, say a transtibial socket or a custom AFO. Produce half through a scan-to-CAD-to-print pipeline with minimal technician input, and half the usual way. Then have clinicians who do not know which is which rate fit, comfort, skin condition and the number of return adjustment visits over three months. Until something like that exists, claims about parity are guesses. You can read how parity is graded for the full scale.
When this could change
Most likely after 2046 (8 in 10 of our scenarios). What that window measures is explained on the replacement-year method page.
Two things could pull it earlier. Cheap scanners and desktop printers keep spreading through orthotic and prosthetic labs, so the digital share of fabrication grows without anyone buying a robot. And robot hands are improving on the exact problem this job poses: handling soft, floppy and sticky materials rather than rigid parts. The robotics tier shown above is a dexterous humanoid, which is the hardest class to deliver and the most actively funded.
Two things hold it back. Most of this job is physical work performed in the presence of a patient, so software gains cap out quickly. And custom medical devices sit inside clinical responsibility and device regulation, which means a named person signs for the fit. Add the capital cost of per-bench hardware against a modest wage bill and the math for a small occupation is slow.
What to do: learn the scan-to-CAD-to-print chain now, because the labs adopting it are hiring the technicians who can run both the digital file and the grinder.
How to stay needed
Lean into the parts of the job that happen with a patient in front of you. Three are worth protecting: taking impressions and measurements directly from the body, fitting and adjusting devices after a wear trial, and repairing or modifying appliances that come back after months of real use. None of these produce clean data for a model, and all of them decide whether the device gets worn.
Two skills pay. The first is digital fabrication: 3D scanning, CAD modification, printer and carver setup, and knowing when a printed part is weaker than a laminated one. The second is clinical communication, meaning the ability to talk with prosthetists, orthotists and patients about what hurts and why, and to translate that into a change at the bench.
Close trades score on similar lines for similar reasons. Compare your own work with Dental Laboratory Technicians, Ophthalmic Laboratory Technicians and Medical Equipment Repairers, or put any two side by side on the job comparison tool. For the wider picture, this job sits in the other production occupations family and is counted under the healthcare sector, and it appears alongside other hands-on trades in the list of jobs that mostly need a person. Every score on this page is built from open data using the published scoring method.