Why lock work stays with people
A locksmith’s day is spent at a door, a safe, or a car. Someone is locked out of a house at 11 p.m. The key snapped in a mortise lock. A business owner needs a safe opened without destroying the contents. That work is physical, improvised and done once, in one place, on hardware that may be 40 years old and poorly documented.
Two tasks make the point. Opening locks and safes when keys are lost or combinations are forgotten depends on feel: reading a plug, listening to a wheel pack, deciding when to drill and where. Cutting and duplicating keys, then fitting and adjusting the cylinder, is the same story in miniature. Software can describe both. It cannot hold the pick, mount the rig, or take responsibility for a hole in someone’s safe door.
Our data also marks this work as heavily physical. Roughly 92% of the task time is hands-on, and the robot class that would be needed to do it is the hardest one we track: a dexterous humanoid. That is why the answer to "will AI replace locksmiths" looks different from the answer for desk-based trades support roles.
What AI does, what it helps with, and what people keep
Start with the part AI can run on its own. Share of task time in that group: 0%. It covers the office end of a locksmith business rather than the van: writing up service records and invoices, drafting quotes, and answering routine inquiries about pricing and coverage areas. None of it touches a cylinder.
Next, the assisted work. Share of task time where AI helps a person rather than replacing them: 8%. Diagnosis and parts lookup fit here. Identifying an unfamiliar lock or safe model, finding the right blank, pulling up a code series or a wiring diagram for an electronic access control board, and scheduling a route across a city all go faster with good software. The judgment call at the door is still yours. Our coverage method explains how we split task time this way.
Everything else needs a person on site: 92% of task time. That is the bulk of the trade. Installing and repairing locks, safes and vault doors. Non-destructive entry and lock changes after a break-in or an eviction. Programming automotive transponder keys and fobs to a specific vehicle. Repairing a damaged safe body or relocking device. Work that mixes tools, trust and liability does not move to software quietly.
What the evidence actually shows
Nothing has tested AI head to head against a working locksmith. Our evidence grade for "is it better than a person?" is D, and that grade means not measured. So we publish no parity number for this job, and you should treat anyone who gives you one with suspicion.
What would settle it is specific: a benchmark where a system, with whatever hardware it likes, opens a set of common residential and commercial locks without damage; a trial of automotive key programming across several makes and model years; and a timed comparison on safe deposit and vault servicing against a qualified technician. Until something like that is published and dated, the honest position is an open question. How we grade evidence is set out in our methodology.
The labor market numbers point elsewhere. The BLS counts about 15,040 locksmiths and safe repairers in the US and a median wage of $51,320 (BLS, 2025). It also projects employment falling 11.7% between 2025 and 2035 (BLS, 2025). That decline is about product change, not chatbots: keyless entry, dealer-controlled car keys, cheaper replace-don’t-repair hardware, and fewer people learning the trade. Demand for emergency and commercial work holds up better than key duplication.
When the picture could change
Most likely after 2046 (8 in 10 of our scenarios). Read that window against our replacement-year method rather than as a prediction about any one shop.
Two things could pull it closer. General-purpose robot arms getting cheap and reliable enough to mount at a door would matter most, since the physical share of this job is where the difficulty sits. So would manufacturers shifting to fully electronic, remotely managed locks that are reset from a dashboard instead of serviced by hand.
Two things push it out. The hardware bill is the first: a dexterous humanoid capable of non-destructive entry does not exist as a product you can hire. The second is accountability. Opening someone’s safe or rekeying a building after a dispute involves identity checks, licensing in many states, and insurance. Our cost comparison for this job runs $10 to $1,120 on the AI side against $1,920 to $4,290 for the human one, and the cheap end buys software for the office, not an engineer at a door. If you want to see how different trades land, put two of them side by side on our compare tool.
How to stay needed in this trade
What to do: move toward the jobs that pay for judgment and liability, not the ones that pay for a key blank.
Three tasks to lean into. First, safe and vault work: opening, repair, relocker replacement and moves. Few people do it well, and the call-out rate reflects that. Second, electronic access control: fitting, wiring and servicing card readers, maglocks and master key systems for commercial buildings, including the service contracts that follow. Third, automotive work: transponder and proximity key programming, all-keys-lost jobs, and ignition repair across makes.
Two skills carry the most weight. One is systems work with low-voltage electronics and networked locks, because the growth in this field is in devices that have both a keyway and an IP address. The other is customer handling under stress, including identity verification and clear records, since that is what insurers and landlords are buying.
Close trades worth a look: Mechanical Door Repairers, Security and Fire Alarm Systems Installers, and Watch and Clock Repairers. The wider group sits on our other installation, maintenance and repair family page, and the business context is covered under the other services sector. For hands-on jobs generally, our list of jobs that mostly need a person and the guide to AI and trades careers are the next useful reads.