AI in Manufacturing

No vendor-pitch hype. Six manufacturing AI plays that consistently survive the shop floor — predictive maintenance, quality vision, process optimisation, plant-floor supply chain, safety AI, and pragmatic digital twin — and three pitches that consistently disappoint when deployed in real plants, named explicitly. The OT/IT bridge filter that sequences your specific plant. Predictive maintenance with sensor coverage decisions and the narrative layer maintenance engineers actually act on. Quality vision with the operator labelling loop and the false-confidence discipline. Process optimisation with operator-led experiments and overfitting guards. Plant-floor supply chain integration with real-time re-planning cadences. Safety AI with the support-not-surveillance line and the one line nobody crosses. Pragmatic digital twin within six-month bounded scope, not eighteen-month megaproject. And a one-page 12-month plant roadmap with OEE and quality KPIs the plant manager owns — not AI metrics nobody on the line recognises.