EXPOSED
Food batchmakers weigh, mix, cook and blend ingredients to fixed formulas, monitor kettle and mixer gauges, and record batch data — work that language AI can't do directly but that industrial process automation has been eating for decades, with sensors and PLCs now handling dosing, temperature holds, and batch logging. The physical handling of ingredients, hoses, filters, and messy cleanup in wet, hot plant conditions is the real moat, and it's a moat against robots, not chatbots. No personal licensure and no customer relationship; discretion is mostly limited to sensory checks and adjusting a batch that's off-spec.
Mixed — a routine tier and a judgment tier. Automated dosing systems, jacketed kettles with PLC temperature holds, and inline batch logging already cover the weigh-mix-cook-record sequence in large plants, but hand-charging minor ingredients from sacks and totes, swapping filters, hooking transfer hoses, and hand-tasting a batch keep this at 11 rather than down near the fully-scripted assembly jobs.
Hands-on in uncontrolled environments. The shift is spent on a wet, hot production floor lifting 50-lb ingredient bags, climbing to mixer hatches, dragging CIP hoses, and scraping out kettles between runs — uncontrolled enough to sit at 13, though it's still one fixed plant with fixed equipment rather than a changing outdoor site.
No licence, no signature requirement. No state licence attaches to the batchmaker; the plant's HACCP plan, the PCQI, and the QA lab carry the food-safety accountability, and the 3 reflects only the food handler card or in-house GMP/allergen training the employer requires.
Anonymous artifact production. The batch leaves as an anonymous lot code on a pallet — no buyer, retailer, or consumer ever knows who ran the mixer, and any relationship is with the supervisor and the next shift, not a customer.
Executes defined procedures on defined inputs. Formulas, cook times, and hold temperatures come off a spec sheet with deviations escalated to QA, so the 6 covers the real but narrow calls — judging viscosity or colour by eye, deciding a batch needs another minute or a rework — inside a documented procedure someone else owns.
Has AI actually changed your work?