IPC International Community magazine an association member publication
Issue link: https://iconnect007.uberflip.com/i/1546580
14 I-CONNECT007 MAGAZINE I SEPTEMBER 2026 T H E R I G H T A P P ROAC H statistical distinction between "chance cause" (com- mon) variation and "assignable cause" (special) variation that Deming would later carry forward and popularize. Shewhart's real achievement was insisting that any claim about a process being "in control" rests on a defensible statistical basis, not intuition or a manager's gut feel. He also originated the Plan-Do- Study-Act cycle that Deming would popularize as the PDCA cycle. Shewhart would likely be the most rigorous skeptic of AI. A model that outputs an anomaly score is, in his terms, an assignable-cause detector, but unlike a control chart built on known statistical distributions, most machine learning models don't come with a transparent, auditable basis for why a given score crossed a threshold. He would push hard for treating model validation with the same seriousness he brought to control limits: back-testing against historical data, quanti- fying false-positive and false-negative rates the way a control chart's three-sigma limits quantify the odds of a false alarm, and refusing to treat a model's output as authoritative until its statistical behavior is understood. For Shewhart, the danger of AI wouldn't be the technology itself, but rather organizations trusting a black box the same way they once trusted an unproven chart without doing the scientific work to justify that trust. "Explainable AI," in his framing, is the modern equivalent of the statistical theory that made the control chart trustworthy in the first place. Deming: AI as the Ultimate Variation Detector Deming's entire framework rested on the insight that most quality problems come from variation in the system, not from the people working in it. His statistical process control charts existed to separate "common cause" variation, the noise inherent in any process, from "special cause" variation that signals something has actually changed. He was fiercely opposed to tampering, meaning adjusting a stable process in response to normal noise, which only makes performance worse. Deming would almost certainly see machine learning as an extension of the control chart, not a replacement for it. A model trained on process data can detect subtle multivariate shifts that a single- variable Shewhart chart would miss, with several parameters drifting together in a way no individual chart would flag. But he would apply the same discipline to AI outputs that he applied to control limits: an anomaly score is not, by itself, permission to act. Any AI-flagged deviation still has to run through the same root-cause discipline before anyone touches the process, because a model that "detects" noise and triggers unnecessary adjustment is just tampering with better technology behind it. His 14 Points would likely gain a 15th in spirit: Understand your model as thoroughly as you understand your process, or you've automated the guesswork you were trying to eliminate. Juran: AI as a Prioritization Engine Juran's contribution was less about statistics and more about management focus, the Quality Trilogy of planning, control, and improvement, and his ad- aptation of the Pareto principle into the "vital few and useful many." For Juran, the central question was always where to spend limited attention, and which a handful of causes account for most of the defects, cost, and customer complaints. AI is, in a very literal sense, a Pareto engine at scale. A model that ingests warranty claims, inspec- tion records, supplier data, and field failures simulta- neously can rank root causes by actual financial and quality impact far faster than a manual Pareto analysis, and continuously rather than as a quarterly exercise. Juran would likely push organizations to use AI first for the unglamorous work of cost-of-quality analysis, accurately attributing the financial impact of failures so that improvement projects get funded on evidence rather than on institutional politics. He'd also warn that a prioritization engine is only as trustworthy as the data governance behind it; garbage inputs produce a beautifully ranked list of the wrong problems. Crosby: AI in Service of Zero Defects Crosby's "Zero Defects" philosophy was as much cultural as technical, and he is famous for his claim that quality is free because the cost of prevention is always lower than the combined cost of failure, appraisal, and rework. He was deeply skeptical of