The huge gap between IoT and AI is not technology. It is training data. In manufacturing, AI for defect reduction requires more than collecting equipment data. Manufacturing data must be aggregated at the batch or lot level and linked to quality inspection results before meaningful training data can be created. Bridging this gap is often the key to successful AI-driven quality improvement and defect reduction.

Many factories visualize defect data, yet the numbers often don’t make sense. In several plants I supported, the same production batch showed different defect rates depending on the inspector—2% by an expert versus 8% by others. The issue wasn’t product quality but human behavior shaped by inadequate KPIs. When inspectors are evaluated by “how many items they check,” speed is prioritized, judgment becomes inconsistent, and defect data becomes unreliable. This leads not only t