Why this work stays in human hands
The job happens at a bench, not on a screen. Mixing and tempering molding material to the right feel, packing it around a pattern, then trimming, patching and smoothing the finished piece are all judgment calls made with hands and simple tools. A model can read a drawing and suggest settings. It cannot tell by touch that the plaster is too wet or that a mold needs a few more taps before it will release cleanly.
The second reason is variety. Much of this work is short runs, repairs, custom patterns and one-off pieces for ceramics, concrete, glass products and stone. Fixed automation pays off when thousands of identical parts come off the same line. Our robotics read for this occupation puts it in the fixed automation tier, with almost all task time physical. That is the kind of setup that replaces a dedicated station, not a worker who moves between patterns, molds and finishing all day.
Demand matters too. The Bureau of Labor Statistics counted about 33,190 people in this occupation with median pay of $46,170 a year, and projects employment growth of 5.8% from 2025 to 2035 (BLS, 2025). That is a small trade, but not a shrinking one. The honest story here is tasks shifting, not the work disappearing.
What AI does, what it helps with, and what people keep
The share of task time our scoring marks as work AI can handle on its own is 0%. That slice sits at the paperwork edges of the job: logging output, keeping production records straight, comparing a spec sheet against what was ordered. None of it touches the mold.
Assistance is a bigger story. Camera systems paired with trained models are good at spotting cracks, voids and surface defects, and simulation software can narrow down material and cure settings before anyone mixes a batch. Our augmentation share reads 5%. Think of it as a second set of eyes on inspection and a faster first guess on parameters.
What stays with people is the bulk of it: 95% of task time. Pouring and packing material, breaking molds apart without damaging the piece, repairing a worn pattern, and hand finishing a casting that came out slightly off all need a person in the room. Our coverage figure, which measures how much task time AI can handle today, reads 6 for this job; you can read how that number is built on the Can AI do it page.
How strong is the evidence
Thin, and we say so. The quality-parity grade here is D. A D grade means there is no published test of an AI system against a trained molder or caster on this job’s real tasks, so we give no parity number at all. Guesses dressed up as scores are worse than a blank.
What would settle it is specific: a timed, blind comparison of a machine cell against experienced workers on short-run molding and hand finishing, measured on scrap rate, dimensional accuracy and rework, with mold changeover included rather than excluded. Published defect-detection accuracy for vision systems on cast parts would help too, since inspection is where the current gains are real. Our full approach, including how grades are set, is on the methodology page.
Good to know: cost is part of the picture, and the physical side of this job is where automation gets expensive fast, because each new pattern or mold shape can mean new tooling.
When this could change
Most likely after 2046 (8 in 10 of our scenarios). The replacement-year method explains what that window covers and how the scenarios are run.
Two things could pull it earlier. Cheaper general-purpose robot arms with decent force control would make small-batch mold handling worth automating, and a plant that standardizes its product line can turn varied work into repeat work that fixed automation already handles well. Our guide to robots and physical jobs covers how slowly that hardware has actually moved.
Two things hold it back. Tooling and integration costs dominate in a trade with modest pay, so the payback math rarely works for short runs. And the tacit part of the skill, knowing how a material behaves on a humid day or why a mold is sticking, is not written down anywhere a model could learn it from.
How to stay needed
Lean into the parts of this job that machines struggle with. Mold and pattern repair is the clearest one: fixing a worn or damaged mold keeps a line running and is rarely a repeat problem. Hand finishing and salvage work is the second, because a part that came out imperfect still has to be judged and saved by someone. Setup and changeover is the third, especially moving between patterns and materials quickly.
Two skills are worth adding. First, running and troubleshooting the automated equipment in your shop, including vision inspection stations, so you are the person who keeps the cell producing. Second, reading drawings and basic CAD, which puts you closer to pattern and mold design work. The guide to AI and the trades goes into how those two skills tend to pay off together.
If you want to see where nearby trades land, look at Foundry Mold and Coremakers, Pourers and Casters, Metal and Potters, Manufacturing. You can also browse the rest of the other production occupations family or the wider manufacturing sector, and see which hands-on roles hold up best on our list of jobs that mostly need a person.
This job’s headline Still needs a human figure is 85 out of 100 (higher is safer). To weigh a move, put this trade and another one side by side on the compare tool.