Beyond six sigma: The next operating model for quality

Philip Crosby’s proposition that ‘quality is free’ rested on a simple economic truth: prevention costs less than failure, rework, scrap, warranty, inspection, and nonconformance. That logic remains sound. But prevention itself must now evolve.

In the AI era, quality will remain ‘free’ only if prevention becomes faster, more connected, and more precise. That means moving beyond isolated automation or proof-of-concept pilots toward a new operating model for quality—one built on four reinforcing capabilities. 

1. Make right-first-time engineering scalable

‘Right first time’ can no longer remain a mindset alone. It must become a more reliable outcome—supported by data, digital continuity, and AI-augmented engineering judgment.

Traditional tools such as DFMEA (Design Failure Mode and Effects Analysis) and PFMEA (Process Failure Mode and Effects Analysis) remain essential. They bring structure to risk identification and prevention. But they are also labor-intensive, document-heavy, and dependent on human recall and expert consistency. AI can help by expanding the memory of the quality system. Instead of beginning every risk review from a blank page, teams can draw from prior failure modes, causes, effects, risk patterns, field returns, supplier escapes, reliability evidence, and design history.

This is not about replacing experts. It is about giving experts a richer context. AI can surface similar historical cases, highlight gaps in detection coverage, suggest causes for review, and support more consistent scoring logic. Engineers still decide—but they decide with better evidence.

At Schneider Electric, this shift is visible in AI-enabled support. Product Bills of Materials and Manufacturing Bills of Materials are analyzed early to identify inconsistencies, missing information, and potential design-to-manufacturing gaps before industrialization.

The cost implication is important: smarter risk identification can reduce both defects and over-control. Quality effort can be focused on what truly matters, rather than everything that can be measured.

2. Move nonconformity from documentation to diagnosis

In many organizations, nonconformity management remains too administrative: detect, classify, contain, document, close. The deeper opportunity is to turn every nonconformity into a structured learning event.

Methods such as 8D, A3, DMAIC, fault trees, and cause-and-effect analysis still provide essential rigor. AI does not replace them. It accelerates them. It can cluster similar failures, identify recurring signatures, suggest likely root-cause families, correlate escapes with process windows, and reveal hidden interactions across parameters.

Schneider Electric’s AI-powered Issue to Prevention assistant reflects this direction. It uses semantic search and natural-language interaction to help teams identify similar issues, uncover lessons learned, and access related root causes, corrective actions, and preventive measures from historical quality data.

The goal is clear: compress the time between signal, diagnosis, and corrective action. The nonconformity record should not be the end of a workflow. It should be the beginning of a learning system.

3. Close the loop from field and factory back to design

Most quality organizations do not suffer from a lack of lessons. They suffer from a lack of continuity. Field issues, plant nonconformities, customer complaints, service notes, and supplier escapes often remain trapped in separate systems or functions.

Closed-loop quality changes that. It ensures that what is learned in the field and factory flows back into product definition, risk assessments, process standards, requirements, design rules, and future FMEA cycles.

AI can accelerate this loop by transforming unstructured complaints, service records, logs, and incident narratives into structured quality knowledge. Events can be classified, summarized, and mapped back to failure modes, CTQs (Critical to Quality), process steps, or requirements.

Schneider Electric’s PFMEA Look-Across concept illustrates the model: when a customer issue is identified, learning from root-cause investigations and 8D actions can be connected to potentially impacted PFMEAs across products, sites, and manufacturing processes. The result is faster review, stronger risk assessment, and better prevention.

This is how corrective quality becomes preventive quality.

4. Detect weak signals before they become failures

The next frontier is predictive quality. Traditional systems are often optimized for visible failures—scrap, defects, returns, warranty events, and nonconformities. But once these signals are visible, the cost has already been incurred.

AI and digital-twin thinking can help detect weak signals earlier. By connecting process conditions, field stress, component behavior, environmental exposure, software interaction, and reliability assumptions, quality teams can build a more dynamic view of risk.

Schneider Electric is investing in a Weak Signals platform that uses AI and advanced analytics to detect early indicators of quality, reliability, safety, and customer-experience issues. It analyzes data from customer feedback, engineering reports, service records, manufacturing systems, and technical documentation to identify subtle patterns that traditional reporting may miss.

The shift is profound: from finding what went wrong to detecting what is beginning to go wrong.