Why the molds still get changed by hand
This job happens at a machine, in a shop, with hot metal or melted plastic moving through it. An operator sets up the mold or die, dials in temperature, pressure and cycle time, runs the first shots and checks them. Then the day turns into watching, adjusting and fixing. None of that is a document task.
Software is good at the parts that look like data: cycle logs, scrap counts, production records, maintenance schedules. It is far weaker at the parts that involve a jammed part, a short shot, a mold that needs cleaning and lubricating, or a machine that starts running hot halfway through a run. Those need hands, eyes and judgment in the same room as the machine.
The economics point the same way. Most of the task time here is physical, so the robotics panel on this page lands on mobile robots rather than on a software subscription. A camera can grade a part. It cannot swap a die, chase a leak or decide that this batch of resin is behaving differently from last week’s.
What AI runs, what it assists, and what the operator keeps
Start with the work software can already handle on its own (0% of task time). It is the paperwork around the machine: recording output and defect counts, tracking cycle times and downtime, and feeding that into production reports. Shops have been automating this since long before current AI tools, and the gain is real but narrow. It explains why our coverage score for this job stays where it does.
Next, the assisted group (10% of task time). Machine-vision inspection can flag flash, warpage or color drift faster than a tired eye at hour nine. Control software can suggest a pressure or temperature change when a parameter drifts. In both cases a person signs off, because a false reject costs money and a missed defect costs more.
The rest stays with people (90% of task time). Setting up and changing molds, loading material, trimming and removing finished parts, cleaning and lubricating tooling, and troubleshooting a machine mid-run are all hands-on. That is the floor under the score, and it moves slowly.
What has actually been tested
Nothing has been tested head-to-head against people in this job. Our evidence grade here is D, which means we have no direct, published comparison of an AI or robotic system against a qualified operator on this occupation’s tasks. So we publish no parity number for it, and no one should quote one.
What would settle it is specific: a measured trial of an automated cell handling mold changeover and first-article setup across different part geometries, with scrap rates, downtime and changeover times reported against a trained operator on the same machines. Vision-inspection benchmarks alone would not do it, because inspection is only one slice of the day. Until that exists, the honest answer is uncertainty, not confidence. How we grade parity evidence explains why a D stays blank rather than guessed.
The labor data is clearer. BLS counts about 150,470 of these jobs in the United States, with median pay near $44,350, and projects employment falling 3.4% between 2025 and 2035 (BLS, 2025). That is erosion, not disappearance: fewer machines tended per shift as lines consolidate, and fewer entry-level openings on the floor.
When the picture could change
Most likely after 2046 (8 in 10 of our scenarios). Read that window alongside the replacement-year method, which sets out what the range covers.
Two things could pull it earlier. Cheaper mobile and arm robotics would make automated part removal and material loading viable for smaller shops, not just high-volume plants. And standardized tooling, where molds and quick-change systems are designed for machines to swap, removes the fiddliest barrier in one step.
Two things hold it back. Capital cost and payback: a new cell competes with an operator’s wage over years, and short runs with frequent changeovers rarely pencil out. And variability: resins, alloys, mold wear and ambient conditions all shift, so a cell tuned for one part often fails on the next. Add plant safety rules around hot metal and press areas, and retrofits get slower still.
How to stay needed on the floor
Lean into the work that sits on the human side of this page’s task split. Setup and changeover is the first: the faster and cleaner you can bring a new mold to a good first article, the harder you are to design out. Troubleshooting is the second: being the person who diagnoses a sink mark, a flash pattern or an intermittent cycle fault is a skill that does not transfer to a dashboard. Tooling care is the third, because mold cleaning, lubrication and early wear spotting protect the most expensive asset in the building.
Two skills raise the floor further. Learn to read and edit machine controls and PLC parameters rather than just pressing recall. And learn basic quality method, including measurement, SPC charts and root-cause work, so you can argue a case with data when a vision system and a customer disagree.
What to do: ask to be trained on the changeover and inspection-system setup at your plant, not just on running the cycle.
If you are weighing a move, nearby work is worth a look: Foundry Mold and Coremakers, Pourers and Casters, Metal, and Extruding and Drawing Machine Setters, Operators, and Tenders. You can put any two of them side by side on our compare tool, see the wider metal and plastic worker family, or read how the whole sector looks in manufacturing. For the physical side of the question, our guide to robots and physical jobs covers what hardware can and cannot do yet, and our method shows how every figure on this page is built.