The AI landscape doesn't move in one direction — it lurches. Some techniques leap from experiment to table stakes in a single quarter; others stall against regulatory walls, technical ceilings, or organisational inertia that no amount of hype can dislodge. Knowing which is which is the hard part. The State of Play cuts through the noise with a rigorously maintained index of AI techniques across every major business domain — classified by maturity, evidenced by real-world adoption, and updated daily so you always know where you stand relative to the field. Stop guessing. Start knowing.
A daily newsletter distilling the past two weeks of movement in a domain or two — delivered to your inbox while the index updates in the background.
Each dot marks the weighted maturity of practices within a domain — hover for a brief summary, click for more detail
AI that monitors business processes for quality deviations and implements statistical process control measures. Includes SPC chart automation and quality alert generation; distinct from quality inspection in manufacturing which checks physical products rather than business processes. Scope covers AI/ML-enhanced quality monitoring and anomaly detection; traditional SPC and Six Sigma methods without ML are out of scope.
AI-enhanced quality management and process control shows accelerating real-world deployment but remains research-stage due to persistent scaling barriers. The practice encompasses automated defect detection, SPC charting, anomaly detection, and quality alert routing. The technical foundation is solid: Leading AI vision systems achieve 99%+ defect detection accuracy; vendor platforms (SAP Joule, Minitab, InfinityQS) are mature; and April 2026 evidence shows quality control and visual inspection as the #1 deployed AI application across manufacturing (53% adoption). Named customer deployments demonstrate clear ROI: North American PCB manufacturers achieving real-time edge processing; Saudi packaging company realizing SAR 3.5M annual savings with 99.4% accuracy; automotive plants increasing vision cameras 30-40x. Yet adoption remains bottlenecked at scale: only 23% of European enterprises with AI pilots reach production scale; false-positive-driven operator distrust erodes deployed systems; OT/IT data fragmentation causes 85% of manufacturing AI projects to fail. The tension has sharpened: concrete post-deployment evidence now exists alongside persistent organizational barriers. Payback periods (6-12 months typical) and ROI multiples (10-25x for predictive maintenance) are attractive, but execution gaps — data governance, ops readiness, change management — still dominate tier classification.
Deployments are scaling, but adoption remains uneven and fragile. Named production deployments confirm the capability-to-ROI path: North American electronics manufacturers deploying edge GPU-based real-time inspection; a Saudi packaging producer (2M+ units/week across 3 lines) achieving 99.4% accuracy and SAR 3.5M annual savings; automotive plants using 30-40x more vision cameras for defect detection; Volvo Group casting facility using predictive ML to prevent defects upstream. April 2026 surveys show quality control as the #1 deployed AI application (53%), with 88% of manufacturers using AI in at least one function and 28% already realizing ROI. Typical economics are attractive: detection rates improve from 75-80% to 99%+, false reject rates <0.5%, payback 6-12 months, labor savings 60-80%. The market itself is expanding: AI manufacturing quality control projected to reach $124.3B by 2034 (22.2% CAGR from $17.1B in 2026). However, deployment-to-scale remains the critical gap. European survey data shows only 23% of manufacturing enterprises with AI pilots reach production scale; Cisco reports only 20% have scaled AI across operations; manufacturing analyst data shows only 12% of AI initiatives are fully deployed, with 37% feeling operationally ready. The critical failure mode: false-positive pseudo-defects eroding operator trust — even small false-positive rates at production volume are sufficient to cause abandonments. OT/IT data fragmentation (sensor data siloed from ERP/MES) causes 85% of quality monitoring dashboards to fail at scale. Organizations spend 90% of data science effort on data janitor work, not model improvement. The practice represents a narrow but widening band of viable deployments (specific use cases, mature vendors, committed customers) surrounded by a vast majority of stalled pilots and abandoned implementations.
— Sandia National Laboratories implements AI-assisted visual inspection in production with human-in-the-loop quality management, demonstrating government-scale deployment of AI quality control systems.
— Production SaaS platform for acoustic anomaly detection in manufacturing quality control, with named customer deployments and documented use cases in assembly and test environments.
— AI-driven quality assurance in additive manufacturing with in-situ monitoring, closed-loop control, and born-qualified parts verification; named organizations and production implementations demonstrating quality process automation.
— Multi-angle inline AI inspection at production speed (1,200 pieces/min, 50MP, 32 illumination channels) with closed-loop MES feedback, real-time root cause analysis, and continuous model retraining; targets FA=0%, FR≤1%.
— FDA warning letter to Purolea Cosmetics establishes first cGMP violation for AI overreliance, documenting regulatory precedent that AI cannot substitute for Quality Unit accountability in pharmaceutical manufacturing.
— Practical deployment case demonstrating computer vision achieving 98%+ defect detection vs 75–80% for human inspectors, with concrete ROI analysis and documented challenges in electronics manufacturing.
— Comprehensive deployment metrics for AI vision systems with specific accuracy, latency, and ROI data across multiple defect types and industries; documents standard technical integration patterns.
— Market and deployment guide covering AI quality control with named case studies (BMW 99.6% accuracy, Foxconn 60% escape reduction) and market sizing ($5.1B in 2025, $68.4B projected by 2032).
2018: Early recognition that traditional SPC methods are insufficient for modern low-volume, high-mix manufacturing; AI experimentation in manufacturing broad but adoption at scale remains at 2%; quality-specific AI-driven process control still emerging without documented large-scale deployments.
2019: Traditional SPC remains misapplied despite established best practices; isolated deployments of automated process control systems (EROWA, smart factory platforms) emerge but no category-wide adoption; Gartner identifies significant underutilization of existing MES and process control capabilities in manufacturing; SAP and enterprise vendors advance AI roadmaps but quality-specific functionality remains undifferentiated.
2021: Vendors release modernized SPC solutions with real-time automation (Minitab Real-Time SPC, August 2021); 80% of manufacturers identify smart manufacturing as strategic priority, with pandemic accelerating smart technology adoption; concrete deployments continue in precision manufacturing (pharmaceutical parts); however, clear AI/ML-driven process control solutions remain scarce, market dominated by incremental SPC tool upgrades rather than AI-enhanced anomaly detection or predictive process control.
2022-H1: Enterprise quality management modules see real adoption (SAP Quality Awards, 14+ named deployments across automotive and industrial sectors); smart manufacturing adoption accelerates 50% year-over-year with 83% of manufacturers viewing as critical; AWS-SAP integrations demonstrate ecosystem maturity for AI-enhanced quality monitoring. However, NSF data shows only 6-7% AI adoption in manufacturing sectors; alert fatigue emerges as critical limitation (43-55% of organizations missing critical alerts due to false positives); practitioners argue traditional SPC is obsolete for real-time monitoring. Practice remains bottlenecked by SME capability gaps, alert system reliability, and lack of proven AI-driven solutions at scale.
2022-H2: Real-world SPC automation deployments begin to appear (Minitab Connect at Toyota assembly plant with API-driven 90-second data polling and automated alert dashboards). APAC manufacturers show strongest momentum toward smart manufacturing adoption (75% adoption by end-2022, 93% viewing as strategic). However, ISG survey reveals persistent implementation challenges: 73% of manufacturers have <2 years experience with smart manufacturing and 70% report minimal progress despite stated priority. Alert fatigue remains critical blocker, with process industry examples citing catastrophic consequences (LNG plant explosion from ignored alarms). Traditional SPC continues showing poor results in real-world deployments (food production case study). Overall signal: adoption intentions strong but execution and alert management challenges constrain the category to research stage.
2023-H1: Enterprise vendors accelerated AI integration: SAP embedded AI into Digital Manufacturing solutions and announced IBM Watson and OpenAI partnerships (April-May 2023). SAP Quality Awards 2023 showed sustained quality management deployments across organizations. Rockwell survey confirmed continued manufacturer appetite for smart manufacturing and AI-driven quality insights. However, WEF analysis exposed critical constraint: 70% of Industry 4.0 and AI pilots fail to move beyond initial stages, indicating severe execution barriers. Fictiv survey confirmed manufacturers' strategic intent to adopt AI but highlighted persistent workforce, economic, and implementation challenges. Vendor platform advancement continues outpacing actual organizational capability to deploy and sustain these systems, keeping the practice in research stage.
2023-H2: SAP reported 24,000+ customers using Business AI across 130+ use cases (November 2023); Deloitte and Accenture launched production services for AI-enhanced quality and supply chain processes. However, Make UK/Infor survey of 135 manufacturers (October 2023) found 55% implementing/planning AI/ML for automation with persistent barriers: skills shortages (46%), data integration challenges (41%), ROI expectations of 5+ years. Quality Magazine (December 2023) cited Gartner data showing 69% of quality leaders piloting predictive analytics but only 17% confident in implementation. Alert fatigue persisted as systemic blocker. Constellation analyst (July 2023) identified cloud migration as critical adoption barrier, with S/4HANA transition delaying AI deployment. Adoption sentiment remained strong but execution challenges and long ROI timelines kept the practice in research stage.
2024-Q1: Manufacturing demand for AI-driven quality management accelerated sharply: Rockwell Automation survey of 1,500+ manufacturers showed 83% planning generative AI deployment in 2024 with quality control as #1 use case. SAP expanded roadmap to 305 total AI scenarios (155 in production, 150 planned for 2024) with 96% of customers having executive AI mandates. However, data quality emerged as primary constraint: 76% of organizations struggled with siloed/low-quality data, costing companies ~6% of annual revenue. Enterprise skepticism remained significant: DSAG survey showed only 28% of SAP users considered AI highly relevant (vs. 65% skeptical). Industry analysis noted AI-powered SPC could reduce false positives by 30%, but implementation barriers persisted. Practice advanced from "strong intent, execution gaps" to "record demand, persistent capability gaps," keeping it in research stage.
2024-Q2: Vendor platforms achieved critical GA milestones (SAP Joule copilot integrated into S/4HANA Cloud and SAP Build; Tricentis AI test automation in SAP Cloud ALM). Academic research demonstrated SQC innovation (ChatSQC combining LLMs with SPC knowledge via RAG). However, adoption paradox emerged sharply: BCG survey of 1,800 execs showed 89% plan AI but only 68% started implementation. Rootstock survey revealed 90% of operators already using AI yet 38% felt they lagged peers—reflecting widespread adoption anxiety despite uptake. Barrier profile shifted to budget (31%) and time (27%) constraints. Practice characterized by rapid vendor momentum and research innovation but persistent confidence gaps between stated and realized adoption; kept in research stage.
2024-Q4: Vendor platform maturity advanced further (SAP Q3 release added AI-assisted visual inspection and process mining; Minitab hardened production APIs for real-time SPC integration with SAP Digital Manufacturing). Enterprise demand remained strong: Rockwell's 1,500+ manufacturer survey showed 95% using/evaluating smart manufacturing, quality as #1 AI/ML use case. However, real-world adoption gaps widened: independent Fraunhofer research documented only 16% German industrial firm AI adoption (30% large enterprises, 13% SMEs); UK Make UK survey showed 36% using AI, 16% knowledgeable. Critical vulnerability emerged: generative AI tools like ChatGPT showed accuracy risks (incorrect SPC phase descriptions, hallucinations) raising trustworthiness concerns. Rockwell data exposed data utilization barrier: 44% of collected data actively used. Practice remained in research stage with advancing vendor capabilities but persistent adoption, regional, and reliability headwinds.
2025-Q1: Demand sustained at elevated levels: Rockwell's March 2025 survey confirmed quality control as #1 AI/ML use case for second consecutive year (50% plan 2025 deployment), with 95% of 1,500+ manufacturers investing/planning AI and 81% citing accelerated digital transformation pressures. Vendor capability expanded: SAP evolved Joule into autonomous "super orchestrator" with expanded AI agent portfolio; analyst reports show AI in half of Q4 cloud deals. However, preparedness barriers remained critical: Riverbed March 2025 survey found only 32% of manufacturers fully prepared despite 92% viewing AI as priority. Academic research (arXiv) provided independent validation of AI/ML methodology maturity (neural networks, LMMs for smart process control), but Deloitte analysis emphasized foundational gaps (data silos, cultural change, cost barriers). Practice remained research-stage with strong intent signals, sustained vendor innovation, and persistent execution/preparedness challenges.
2025-Q2: Deployment outcomes emerged with sustained intent: Deloitte survey (May 2025) of 600 manufacturers reported smart manufacturing driving 20% production output improvement, 20% productivity gain, and 15% capacity unlock, with 92% believing smart manufacturing will drive competitiveness over next 3 years. Rockwell's 10th annual survey (June 2025, 1,500+ manufacturers across 17 countries) showed 56% piloting smart manufacturing, 20% at scale, 20% planning—with 95% having invested or planning AI/ML investment over next 5 years. However, critical scaling barriers persisted: Applied AI analysis documented only 26% of organizations successfully scaling AI beyond pilots despite 72% adopting in at least one function, with 70% of barriers organizational (people, process, change management). HBR expert panel (April 2025) emphasized that business processes must be fundamentally redesigned for AI to deliver value, with AI initiatives frequently failing without process optimization. Practice showed advancing demonstrated ROI alongside persistent execution challenges, maintaining research stage as majority of deployments remained in pilot or early scaling phases.
2025-Q3: Vendor platform and market growth accelerated: Real-Time SPC software market reached $2.35 billion with 7.8% CAGR; SAP Joule evolved into autonomous agent orchestrator; vendor ecosystem deepened partnerships (InfinityQS-SAP, Siemens Opcenter releases). However, productivity and ROI risks surfaced: MIT analysis of Census Bureau data revealed AI adoption J-curve with initial 1.33pp productivity decline before recovery; ISG found only 31% of 1,200 AI use cases in full production with significant ROI underdelivery (25% growth, 50% efficiency). ETQ survey of 752 quality leaders showed 99% AI adoption/planning yet independent case analysis documented major ERP implementation failures ($125-$1,000M) due to change management and user resistance. Practice remained research-stage with the tension sharpening: organizational demand high and vendor capabilities advancing, but real deployments facing productivity headwinds and execution barriers.
2025-Q4: Vendor platform maturity sustained; SAP Joule reached 400+ AI features with ISO 42001 governance; Minitab hardened production APIs; InfinityQS-SAP partnerships deepened. Strategic demand remained elevated: 80% of manufacturing execs planning 20%+ smart manufacturing budgets (Deloitte). However, Q4 brought critical evidence of ROI underdelivery: only 6% of AI projects achieve 1-year ROI with 2-4 year typical payback (Deloitte survey of 1,854 execs); 30% of GenAI projects abandoned after POC (Gartner); up to 95% deliver zero measurable ROI. Pharma manufacturing research found employee acceptance and organizational readiness as critical success factors, not technical capability. Practice remained research-stage as extended payback timelines, high abandonment rates, and organizational barriers constrained real-world deployment despite sustained vendor momentum and strategic intent.
2026-Jan: Enterprise AI tool deployment scaled to 60% of workforce (50% YoY growth), with 85% of major organizations customizing autonomous agents; vendor ecosystems matured (400+ SAP features, production APIs) yet ROI execution gaps widened sharply—PwC found 56% of CEOs see zero ROI, MIT documented 95% failure rate for GenAI projects, positioning 2026 as Trough of Disillusionment inflection where payback timelines and organizational readiness became tier-determining factors.
2026-Feb: SPC tool ecosystem remained mature (Minitab leading 44% of pharma SPC adoptions, mature vendor ecosystem), yet independent analysis documented critical ROI barriers: only 5% of enterprises achieve substantial AI returns with 35% partial returns; alert fatigue persisted as systemic limitation with 90% of monitoring alerts requiring no action; 52% of manufacturers use AI for quality control but data quality (45%) and skills gaps (38%) remain primary barriers to advancement.
2026-Mar: Concrete production deployments documented: BMW deployed Landing AI visual inspection achieving 40% defect escape reduction; pharma manufacturers (Hyperbolic case studies) deployed AI inspection at 500K+ tablets/day with 99.8% accuracy and 80% inspector time reduction with regulatory compliance. Standards matured: AIAG-VDA released first major SPC standard update (2026) explicitly addressing digital/automated SPC as Industry 4.0 requirement. Vendor platforms advanced (NEXSPC 4.0 addressing alert fatigue with customizable rule groups). Market signals sustained: 80% manufacturing execs plan 20%+ smart manufacturing investment (Deloitte 2025); defect detection market projected $6.6B by 2034 from $3.3B in 2024. However, systemic barriers persisted: analysis documented why AI projects fail without factory operational insight (data fragmentation); Wirtek consulting analysis notes 34B-to-155B market growth projection (2025–2030) but technology bottleneck remains organizational readiness. Practice remained research-stage: strong deployment signals and vendor maturity now coupled with explicit standards recognition, but organizational barriers and preparedness gaps still dominate tier classification.
2026-Apr: Survey of 500 global manufacturing professionals confirms AI-powered quality control and visual inspection as the #1 deployed AI application (53%), with 88% using AI in at least one function and 28% realising ROI — the strongest adoption signal to date. Named deployments reinforce ROI: Saudi packaging manufacturer achieves 99.4% accuracy across 3 lines (2M+ units/week) with SAR 3.5M annual savings; North American PCB manufacturers deploy real-time edge GPU inspection on SMT lines; automotive plants scale cameras 30-40x for defect detection. Typical economics now documented: detection improves 75-80% to 99%+, false reject rate <0.5%, 6-12 month payback, 60-80% labor savings. Regulatory enforcement failure surfaces: an FDA warning letter cites a pharma manufacturer for improperly delegating quality and compliance decisions to AI without human validation, documenting governance gaps and sector-specific adoption risk. Grant Thornton's 2026 AI Impact Survey (950 business leaders) reveals a critical execution gap: 62% of manufacturers focus AI on operations yet only 7% have tested incident response, and zero manufacturers report significant cost savings from AI (versus 12% across all industries) — reinforcing ROI underdelivery dynamics. Critical failure dynamics persist: false-positive pseudo-defects erode operator trust; Gartner attributes 85% manufacturing AI failure to OT/IT data divide; only 23% of European enterprises with AI pilots reach production scale. Practice advanced from research to bleeding-edge tier on strength of deployment breadth, but execution barriers — data fragmentation, regulatory governance, operator trust, organisational readiness — remain the binding constraint.
2026-May: Architectural maturity evidence emerged: UnitXLabs documented closed-loop zero-defect manufacturing architecture with inline multi-angle AI inspection (1,200 pieces/min, 50MP, 32 illumination channels), real-time MES feedback, root cause analysis, and continuous model retraining (target: FA=0%, FR≤1%). Sandia National Laboratories deployed AI-assisted visual inspection in production with human-in-the-loop quality management. Additive manufacturing industry documented AI-driven quality assurance with in-situ monitoring and born-qualified parts verification across named organizations. Acoustic anomaly detection SaaS (MHP Sounce) reached production deployment with named customer implementations. Vision system deployment guide (ThreadMoat) documented 98%+ detection rates vs 75–80% baseline, with specific ROI analysis in electronics manufacturing. However, regulatory constraints crystallized: FDA warning letter to Purolea Cosmetics established first formal cGMP violation for AI overreliance, precedent that AI cannot substitute for Quality Unit accountability. Practice tenure at bleeding-edge remains sustained but regulatory scope boundaries and persistent scaling execution barriers prevent advancement toward established.