Why this work stays on the shop floor
Ask will AI replace multiple machine tool setters, and the answer sits in the physical half of the job. These workers set up and run two or more machines at once: a press, a grinder, a saw, a molding machine. The day is spent walking between them, loading stock, pulling finished parts, listening for a change in sound and stopping a machine before it scraps a run. Software can read a drawing. It cannot feel a loose fixture or notice that coolant is spraying the wrong way.
Setup is the hardest part to hand over. Aligning tooling, clamping a workpiece, dialing in feeds and speeds for a specific material, then cutting a first article and measuring it: each step needs hands, eyes and judgment about this machine on this day. Older equipment makes it harder still, because much of it has no sensors to report what is happening inside.
The second anchor is fixing things mid-run. Clearing jams, swapping worn tooling, adjusting for a batch of stock that is slightly off spec. That work is unpredictable, and it happens in a tight space full of chips, oil and moving parts. Our data puts the physical portion of the role at the level where mobile robots, not software alone, would be the deciding technology, and that hardware is slow and costly to install across a shop.
What AI does, what it helps with, and what stays with people
A small share of the tasks can be handled by software with no person in the loop: 0% of task time. That end of the list is paperwork-shaped work, such as recording production counts and run data, and working out dimensions and tolerances from a specification sheet. Both are reading, math and logging.
A second group is where AI assists. Machine vision can flag parts that fall outside tolerance, and monitoring software can watch spindle load or cycle times and raise a flag before a tool breaks. An operator still decides whether to stop the machine, scrap the part or adjust the offset. Those checks get faster with software; they do not get done without someone standing there.
The rest belongs to people: 87% of task time. Setting up and changing over machines, loading and unloading material, clearing jams, lubricating and cleaning equipment, and keeping several machines running without a collision between them. That is the core of the role, and it is why the coverage figure above stays where it is. You can read how we measure that figure on the coverage method page.
What the evidence shows so far
There is no direct head-to-head test of AI against a person doing this job. Our evidence grade reflects that: D. A D grade means not measured, so we publish no parity number for machine tool setters and operators. Anyone quoting a precise risk percentage for this occupation is modeling, not measuring.
What would settle it is specific and testable. A timed changeover trial, where a robotic cell and a trained operator each set up the same job on the same machine, with first-article inspection and scrap rates recorded. A jam-clearing and tool-change test on production equipment rather than a demo rig. A multi-machine tending study showing how many machines a cell can keep fed without a person nearby. Until that work is published and repeatable, the honest position is that the physical tasks are untested at human standard. Our full approach is set out in the methodology.
The labor market numbers are firmer. BLS counts 124,590 people in this occupation in the United States, with median pay of $47,180 and projected employment change of 0.6% from 2025 to 2035 (BLS, 2025). That is close to flat: not a collapse, not growth.
When the picture could change
Most likely after 2046 (8 in 10 of our scenarios). Two things could pull that earlier. Cheaper, more capable mobile robots and standardized robotic tending cells would make multi-machine loading a purchase decision rather than an engineering project. And a wave of new equipment with sensors and open data feeds built in would give software something to act on, instead of a machine that reports nothing.
Two things hold it back. Most shops run mixed, aging equipment in short batches, and retrofitting each machine costs real money and downtime. Safety and quality sign-off is the other brake: someone has to own the first article, the scrap rate and the lockout before a line restarts. For what the range does and does not claim, see how we build the replacement year.
How to stay needed in a machine shop
Lean into the tasks that stay hands-on. First, setup and changeover: the faster and more accurately you can swap a job, the harder you are to route around. Second, in-process problem solving, from jams to tool wear to material that arrives out of spec. Third, running several machines at once, which is the skill the job title is built on and the one that keeps labor cost per part down.
Two skills to add. Learn CNC programming and editing, even at a basic level, so you can adjust offsets and read G-code rather than wait for a programmer. Then learn inspection and measurement properly: micrometers, gauges, CMM basics and reading GD&T. Quality sign-off is where automated inspection still needs a person to make the call.
What to do: ask your employer which machines on your floor have usable sensor data, and volunteer for the setup and inspection work on those cells.
Close neighbors worth comparing are CNC tool operators, lathe and turning machine setters and machinists, which lean further toward skilled setup and programming. You can put any two side by side on the compare tool, see the wider metal and plastic workers family, or read how exposure looks across manufacturing. For the hardware side of the story, our guide to robots and physical jobs covers what machines can and cannot do yet.