Why heat work stays next to the equipment
This job is about controlling heat on purpose, in a specific vessel, with a specific material inside. Operators and tenders light and bank burners, charge and draw loads, watch temperature and pressure, pull samples, and shut things down when a reading looks wrong. Software can hold a set point. It cannot smell a scorched batch, feel a sticky door, or decide that a cracked refractory brick means the kiln comes down today.
Most of the task time here is physical and local. Loading, unloading, cleaning, and clearing jams all happen in hot, dusty space with moving material. The robotics panel on this page puts the hardware you would need at mobile robots, not a fixed arm bolted to a bench. That is a harder and more expensive step than adding a control model to an existing panel, and it is the main reason the question of whether AI will replace furnace, kiln, oven, drier, and kettle operators and tenders does not have a quick answer.
There is also a liability floor. A runaway oven or a kettle boil-over is a safety event, not a bad output file. Plants keep a named person responsible for startup, shutdown, lockout, and the call to stop. That responsibility is hard to hand to a model, even a good one.
What software runs, what it assists, and what stays with the operator
On the part AI can do without a person, the share is 0%. That slice is screen work: holding a recipe against a set point, logging readings, and spotting drift in temperature, feed rate, or pressure earlier than a human watching a trend line. Process-control vendors have sold versions of this for years; machine learning makes the alarms smarter, not the furnace self-tending.
The assisted slice is 19%. Here a model suggests and the operator decides. Think of tuning fuel and air for a drier run, scheduling a cleaning window from wear data, or ranking which alarm matters first during a bad shift. The work still happens, just with a better prompt in front of it.
The share this page scores as needing a person is 81%. That is the hands: charging and drawing material, breaking out clinker or buildup, swapping tooling and gaskets, inspecting the lining, and making the judgment call when the product looks off but the sensors read fine. Our overall coverage figure, meaning the share of task time AI can handle today, is 14 out of 100; you can read how that is built on the coverage method page.
What the evidence actually shows
The parity grade for this job is D. That means there is no direct, published test of an AI system against a qualified operator on this job’s real tasks. No one has run a fair trial where a model tends a kiln or kettle through a full shift and the results are compared with a trained tender on the same line. So this page gives no parity number, and you should treat any site that gives you one for this occupation with care.
What would settle it: a plant-level study of an autonomous control system running a furnace, drier, or kettle across varied feedstock, with yield, energy use, scrap, and safety incidents measured against human-tended runs over months, not a demo week. Published results from a manufacturer or a university lab would move the grade. Until then, the honest reading is that the control layer is well proven and the tending layer is untested. How grades are assigned is set out on the quality parity method page, and the full method is at needsahuman.com/methodology.
The labor market numbers are steadier. BLS counts about 14,280 people in this occupation, with median pay of $48,040 and projected employment change of +2.6% over 2025 to 2035 (BLS, 2025). That is a small, slow-growing occupation, not one in free fall.
When this could change
Most likely after 2046 (8 in 10 of our scenarios). What that window measures, and how we build it, is explained on the replacement year method page.
Two things could pull it earlier. First, cheap mobile robots that can charge, draw, and clear a hot vessel without a custom install would attack the physical share directly. Second, new plant builds designed around closed-loop control from day one skip the retrofit problem entirely, so each new line opens with fewer tending roles than the one it replaces.
Two things hold it back. Existing furnaces and kilns are long-lived capital, often decades old, with sensors and doors that were never designed for a machine to work around. And safety regulation plus insurance keep a qualified person on the floor during firing and shutdown, regardless of how good the control model gets. The gap between software cost and labor cost shown on this page is real, but it only pays off once the hardware around the vessel is replaced too.
What to do: if your plant is adding closed-loop control, ask to be trained on the control system rather than left tending around it.
How to stay needed in this job
Lean into the work that keeps a person on the floor. First, physical intervention: charging, drawing, clearing buildup, and changing worn parts safely under heat. Second, equipment condition judgment, meaning reading refractory wear, door seals, and burner behavior before a sensor catches it. Third, startup and shutdown ownership, including lockout and the authority to stop a run.
Two skills raise your floor. One is process-control literacy: knowing what the model is optimizing, where its data comes from, and when its recommendation is wrong. The other is maintenance and instrumentation basics, so you can calibrate, troubleshoot, and keep the sensors that the control layer depends on honest.
Nearby work worth comparing, if you want options inside the same family: metal refining furnace operators and tenders, chemical equipment operators and tenders, and heat treating equipment setters, operators, and tenders. You can put any two of them side by side on the compare tool, see where they sit in manufacturing, or browse the rest of the other production occupations family. For wider context on how physical work scores, the list of jobs that mostly need a person is a useful next stop. This job’s headline Still needs a human score is 80 out of 100 (higher is safer).