Global Lighthouse Network 2026

Closing the loop: turning customer feedback into autonomous quality improvement As suppliers and manufacturing nodes become more connected, transformation extends beyond operations into customer experience. When faced with challenges in managing customer complaints and quality improvements across global operations, Midea in Si Racha, Thailand, implemented an AI-driven system that automates its end-to-end voice-of-customer (VOC) to voice-of-process (VOP) workflow. Customer feedback is captured through a centralized platform and linked to the quality management system (QMS), forming a closed data loop. Graph-matching algorithms and traceability data identify responsible units, while LLMs infer root causes validated by engineer chatbots and auto- generate corrective action recommendations (Figure 27).51 By connecting customer signals directly to production response, Lighthouses are turning feedback into foresight, closing the loop between data and action. Midea’s quality of VOC (voice of customer) to VOP (voice of process) in seven steps FIGURE 27 redro ksat noitucexe hguorht pool esolC — Morning meeting “one-pager” — Process contr ol centre war ning — Long-term quality pr oject segnahc dradnatS are implemented7eussi ytilauq fo essuac toor 3-pot yfitnedI — Quality issue expert knowledge library — Chatbot interaction to finalize r oot causes eussi ytilauq fo seidemer dnemmoceR esuac tooR analysis5 noitnevretnI recommendations6Classify the category of key components — Market r epair r ecords, historical br eakdowns & quality issues — Multi-lingual translation, intent r ecognition & clusteringtnenopmoc lacitirC rotaropavE :yrogetaC egakael rotaropavE :noitpircsed esuac tluaF Welding equipment nozzle wear and ageing .A B. Operator actions C. W elding materials ro ecnanetniam seriuqer ssecorp gnidlew ehT .A replacement of the welding nozzle every two months gninraw ecnanetniam eudrevo dda ot deeN .B function for err or pr evention and contr ol stneve tcejorP :serusaem mret-gnoL lortnoc ssecorP :serusaem mret-trohS sgniteem gninroM :serusaem mret-trohS eussi ytilibisnopseR tcefed egakaeL :noitacfiissalC ssecorP :noitubirtsid ytilibisnopseR tinu elbisnopseR Workshop: WK01 20001KWCAR :eniL tfihS B :maeT eman noitisoP Welding elbatnuocca ot stnenopmoc yek etaleR root location — Accountable department, plant/line/shift and process step — Decision tr ee logic informed by pr ocedur e, malfunction & accountability dataUser Step ,reenigne selasretfA quality engineer ,rotcepsni ,redael tfihS process engineerreenigne ytilau Q ksir ytilauQ identification1 ksir ssecorP identification2 noitamrfinoC of risk units3 fo noitamrfinoC risky pr ocesses4Enabled by AI Joint-developmentGenAI-enabled customer complaints analytics and closed-loop pr ocess quality impr ovements Quality engineers faced delays r esolving 1,000+ annual complaints fr om 800 service centr es acr oss 10+ countries, with manual pr ocesses taking 60+ days. Limited experience (only 20% with 3+ years) led to reliance on judgement. A knowledge base with 3,000+ expert inputs, AIGC and AI algorithms (e.g. clustering, decision tr ees) automated the seven-step VOC-to-VOP pr ocess, cutting r esolution time fr om 3 months to days. The solution uses NLP to structur e descriptions, analyse r oot causes and r ecommend actions via LLMs and case libraries, driving continuous impr ovement despite limited expertise.1 day eussi ytilauQ to action plan lead time-43% etar tcefeD -32% remotsuC complaints rateMidea Si Racha, Thailand Source: Global Lighthouse Network. Global Lighthouse Network: Rewiring Operations for Resilience and Impact at Scale 41