Plant-floor AI usually fails at one step of one job: a handover brief that drops a lockout still in progress, a BI copilot that answers a downtime question from general knowledge because the data was not in the model, a camera clip used to discipline the person in it when the real problem was a blind corner, an injury report that calls an overnight hospital stay "not reportable" without knowing whether the employee was admitted, a translated work instruction that loses a "do not", or a proficiency score that quietly becomes an employment decision. This course follows one plant through those jobs. Corliss Bend Foods (a fictional composite) is a union-represented co-packer in Joliet, Illinois, whose biggest customer will decide on February 9, 2027 whether to keep its bar volume. Each module solves one of the plant's problems, compares the approaches plant leaders actually choose between in October 2026 (enterprise assistants, MES and connected-worker copilots, Power BI Copilot, camera analytics on existing CCTV, built-in app translation), shows a realistic failure and its fix, and ends with a deliverable you can adapt to your own plant and a Decision Lab with consequences. The course closes with a capstone simulation of the February review and a capstone work product, the plant's AI operating standard, which you score against a published rubric as practice.
Three rules run through every lesson: every number traces to a system of record, every safety and quality item survives the summary, and every people decision stays with a named human. The rules taught are the ones that already govern plant leaders: OSHA Part 1904 recording, reporting and anti-retaliation (1904.7, 1904.39, 1904.35), forklift refresher and evaluation triggers (1910.178(l)), lockout continuity at shift change (1910.147(f)(4)), OSHA's 2010 policy that training must be understood, FMLA's no-fault attendance rule, the Defend Trade Secrets Act's "reasonable measures", Illinois Public Act 103-0804 on AI in employment decisions (effective January 1, 2026), and CISA's guidance on AI in operational technology. Vendor products are named to show the categories in use as of October 2026, not as endorsements, and vendor ROI figures are treated only as claims to test against your own baseline. Planning and OEE depth is in mfg-225, quality in mfg-235 and EHS programs in mfg-245; the individual contributor's view is in mfg-110. Educational content, not legal advice.