Why the shift stays with a person
Extruding, forming, pressing, and compacting machine setters, operators, and tenders work inches from hot metal, plastic, glass, or powder. The job is physical first and digital second. A setter dials in a press or extruder for each run, bolts on the right die or screen, and runs test pieces until the part comes out to spec. A tender keeps the machine fed by hand or conveyor, watches it run, and steps in when it stops. None of that happens through a keyboard.
Then there is the repair end of the day. Operators take equipment apart to swap nozzles, punches, and filters, clear a plug in a barrel, and clean and lubricate parts between runs. Material behaves differently batch to batch. The fix often starts with a sound, a smell, or a change in the color of the melt. A model can log that something shifted. A person has to go find out why, with tools in hand.
Inspection sits in the middle. Camera systems are good at catching an out-of-tolerance dimension on a steady line. Deciding whether to scrap the run, slow the line, or re-set the die is a judgment call with money attached, and it belongs to whoever is standing there. How we weigh that kind of split is set out on the coverage method page.
What software runs, what it assists, and what stays with the operator
Only the screen-shaped parts of this job can run without a person: pulling sensor readings into a report, logging counts and run times, queueing the next order. Share of task time AI can handle on its own: 0%. That work is real, and it is the part a plant automates first, because it needs no hardware on the floor.
Assistance is the more useful story on a shop floor. Share of task time where AI supports the operator: 9%. In practice that looks like process monitoring that flags a drift in pressure or temperature before the parts go bad, vision checks on finished pieces, and maintenance prompts tied to cycle counts. The operator still makes the adjustment.
Everything else needs hands and eyes on the machine. Share of task time that needs a person: 91%. Setup and changeover, feeding and clearing material, and stripping a head down to replace a worn part are all in that column. The robotics panel above shows why: most of this job’s task time is physical, and the robot tier that would be needed is not a bolt-on upgrade.
The evidence, and what is still missing
Our evidence grade for “Is it better than a person?” is D. That grade means no one has published a direct test of AI or robots against trained operators in this occupation, so we give no parity number at all. Exposure estimates and general automation indexes are not the same thing as a measured head-to-head result.
A clean test would settle it. Run the same material to the same tolerance on a live line, then compare an automated cell with a qualified setter on changeover time, scrap rate, and unplanned downtime, and publish the method. Until something like that exists, the honest read is that the data describes the task mix, not a contest. You can see how we grade evidence on how we score every job.
The market side is plainer. The Bureau of Labor Statistics counts about 58,770 of these jobs in the US at a median wage of $45,760 (BLS, 2025), with employment projected to change by roughly 1.5% between 2025 and 2035 (BLS projections, 2025). That is a flat line, not a collapse, and it sits inside a sector under steady pressure from offshoring and capital spending cycles. Wider context is on our manufacturing sector page.
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. First, mobile robots and gripper hardware keep getting cheaper, and the monitoring software layer already costs a fraction of an hour of labor, as the cost panel above shows. Second, new lines get built. When a plant replaces an extrusion or pressing line at a capital refresh, the automation can be designed in from the start, which is far easier than retrofitting a thirty-year-old press.
Two things hold it back. Every plant runs different dies, materials, and machine makes, so an automated cell has to be re-engineered line by line, and the engineering cost does not fall as fast as the hardware. And short runs with frequent changeovers are exactly the pattern robots handle worst. The blockers listed above are mostly of that kind: variation and handling, not missing intelligence.
What to do: ask who gets trained on the new controls and robot cells next time your plant buys a line, and put your name on that list.
How to stay needed in this job
Three parts of the work are worth leaning into. Setup and changeover judgment is the first: the person who can get a new die or mold running to spec in half the usual time is the person a plant protects. Fault diagnosis is the second, because clearing jams and tracing a bad batch to its cause is still read from the machine itself. Quality calls are the third: knowing when to scrap, slow, or adjust saves more money than any report.
Two skills pay on top of that. Learn the control side properly, including HMI screens, recipes, and basic PLC fault codes, so you can tell a sensor problem from a process problem. Then learn to tend and maintain automated equipment, including robot cells and vision checks, so when a line is upgraded you run it instead of watching it arrive.
If you want a nearby move, the closest work sits in the same machine families: extruding and drawing machine setters, molding, coremaking, and casting machine setters, and crushing, grinding, and polishing machine setters. Each has its own task split and its own window.
From here you can put two of them side by side on the job comparison tool, see where this role sits among other production occupations, or read the wider picture in our guide to robots and physical jobs. The jobs that most need a person list shows which work holds up best.