隨著生成式人工智慧（Generative Artificial Intelligence, GenAI）與大型語言模型（Large Language Models, LLMs）技術的快速突破，全球製造業正積極探索如何將其整合至內部營運流程中。面對人口結構變化導致的勞動力短缺與人力成本上升，導入生成式 AI 已成為企業實現「降本增效」與智慧轉型之關鍵策略（Brynjolfsson et al., 2023）。GenAI 具備強大的自然語言理解、知識擷取與自動化內容生成能力（Füller et al., 2022），已被廣泛應用於知識管理與決策支援系統中（Doanh et al., 2023）。然而，如何將其系統化地嵌入特定專業品質流程（如客訴處置與風險預防），並實證其對組織效率與人力賦能之實質效益，仍為極具研究價值之課題。
傳統八大問題解決流程（8 Disciplines Problem Solving, 8D）高度依賴工程師的實務經驗與隱性知識，且客訴調查資料分散於不同系統，導致跨系統查詢耗時，影響調查效率與經驗傳承。為破除此瓶頸，本研究結合生成式 AI 與檢索增強生成（Retrieval-Augmented Generation, RAG）技術，建構一套賦能 8D 流程並與 FMEA（Failure Mode and Effects Analysis）深度結合之 Supervisor 監督式多代理人（Multi-Agent）整合系統。
本研究採用設計科學研究法（Design Science Research Methodology, DSRM）結合單一個案研究法，產出兩項核心模型：以 8D 流程 D1 至 D8 為骨幹之「8D×GenAI 整合流程模型」，以及以結案知識自動回饋為核心之「FMEA 知識回饋模型（品質知識飛輪）」。
實證結果顯示，本系統能顯著提升品質工程師（CQE）之處置效率與報告品質：
工作處置效率：客訴調查與報告草案生成之平均處理時間由傳統人工模式的 1,071 分鐘（約 17.85 小時）大幅壓縮至 87 分鐘（約 1.45 小時），總體時間縮短率達 91%；其中極低頻歷史客訴案例之檢索時間更由 8 小時以上縮短至約 10 分鐘，檢索效率提升約 48 倍。
實務有用性與品質：經 10 位專業品保與製程工程師盲測評估，系統生成內容之平均有用性得分達 7.4 分（滿分 10 分），證實其具備高度實用價值，能有效降低工程師之認知與撰寫負擔。
此外，本研究於系統工程落地的過程中，歸納出三項關鍵設計發現：
術語解析前處理層：企業內部專有名詞與縮寫字會造成 Supervisor 意圖解析障礙，須建置術語解析前處理層以確保任務準確路由。
跨系統與 FMEA 資料標準化：跨產線機台與失效模式定義不一致會干擾跨線比對，須先推動資料標準化與後台預處理，方可發揮 AI 最大效益。
權限分群（Permission Grouping）治理：若採單一 Agent 獨立權限控管會導致 Power User 族群極小化，採用權限分群設計能大幅降低使用門檻，擴大多代理人協作之整體效益。
綜觀而言，本研究成功驗證了 GenAI 多代理人架構於製造業客訴處置之實務可行性，貫徹「人機協作（Human-AI Collaboration）」原則，將分散的客訴經驗與FMEA轉化為永續累積的組織智慧資產。最後，本研究亦針對不同 LLMs 評估、跨品質情境擴展及即時數據流串接等方向提出未來研究建議。

With the rapid advancement of Generative Artificial Intelligence (GenAI) and Large Language Models (LLMs), the global manufacturing industry is aggressively exploring the integration of these technologies into internal operations. Amid labor shortages and escalating human resource costs caused by demographic shifts, deploying GenAI has become a strategic cornerstone for enterprises pursuing cost reduction, efficiency optimization, and intelligent digital transformation (Brynjolfsson et al., 2023). GenAI offers powerful capabilities in natural language understanding, knowledge extraction, and automated content generation (Füller et al., 2022), having achieved widespread adoption across knowledge management and decision support systems (Doanh et al., 2023). However, systematically embedding GenAI into domain-specific quality assurance workflows (e.g., customer complaint handling and risk mitigation) and validating its empirical impact on operational efficiency and human workforce empowerment remain critical research pursuits. Traditional Eight Disciplines (8D) problem-solving relies heavily on engineers' heuristic experience and tacit knowledge, while historical complaint records are frequently siloed across disparate systems, creating substantial retrieval latency and hindering organizational knowledge transfer. To address these limitations, this study combines GenAI and Retrieval-Augmented Generation (RAG) to develop a supervisor-directed multi-agent integrated system that augments the 8D lifecycle with Failure Mode and Effects Analysis (FMEA). Utilizing the Design Science Research Methodology (DSRM) coupled with a single-case study, two core artifacts are formulated: the "8D × GenAI Integrated Process Model," spanning disciplines D1 through D8, and the "FMEA Knowledge Feedback Model (Quality Knowledge Flywheel)," centered on the automated closing-loop feedback of case resolutions. Empirical validation demonstrates significant enhancements in customer quality engineers' (CQEs) productivity and output quality: average case handling and drafting duration was compressed from 1,071 minutes (~17.85 hours) to 87 minutes (~1.45 hours), yielding a 91% time reduction, while search turnaround for ultra-low-frequency cases improved 48-fold (from >8 hours to ~10 minutes). Double-blind evaluations by 10 professional QA and process engineers produced an average utility score of 7.4/10, confirming substantial reductions in cognitive load and documentation effort. In addition, three industrial deployment principles are identified: (1) establishing a terminology preprocessing layer to resolve internal domain jargon and ensure accurate supervisor agent routing; (2) enforcing cross-system and FMEA data standardization to enable valid cross-line comparative analytics; and (3) instituting a permission grouping governance framework to lower usage barriers and maximize multi-agent collaboration. Overall, this research confirms the practical viability of multi-agent GenAI frameworks in manufacturing complaint handling, embodying the human-AI collaboration paradigm and converting fragmented failure experiences and FMEA knowledge into sustainable intellectual assets, while suggesting future avenues in heterogeneous LLM evaluation, multi-scenario quality extension, and real-time streaming integration.

一、中文部分
1. 104人力銀行、工業技術研究院（2025）。《2025半導體業人才報告書》。104人力銀行。
2. 施明欣、陳正芳（2007）。應用知識管理於IC封裝業客訴8D Report案件之探討。品質月刊，43（2），42–46。
二、英文部分
1. AIAG. (2012). CQI-20: Effective problem solving practitioner guide (1st ed.). Automotive Industry Action Group.
2. Asai, A., Sewon, M., Pang, R. Y., Wei, H., Lin, X. V., Koncel-Kedziorski, R., Smith, N. A., & Yih, W. (2024). Self-RAG: Learning to retrieve, generate, and critique through self-reflection. Proceedings of the International Conference on Learning Representations (ICLR 2024).
3. Astanti, R. D., Sutanto, I. C., & Ai, T. J. (2022). Complaint management model of manufacturing products using text mining and potential failure identification. The TQM Journal, 34(6), 2056–2068. https://doi.org/10.1108/TQM-01-2021-0021
4. Baldazzi, T., Bellomarini, L., Cima, L., & Laureti, A. (2023). Fine-tuning large enterprise language models via ontological reasoning. arXiv preprint arXiv:2312.08381.
5. Banica, C.-F., & Belu, N. (2019). Application of 8D methodology – An effective problem solving tool in automotive industry. Scientific Bulletin, Automotive Series, 25(29), 1–7.
6. Behrens, B. A., Wilde, I., & Hoffmann, M. (2007). Complaint management using the extended 8D-method along the automotive supply chain. Production Engineering, 1, 91–95. https://doi.org/10.1007/s11740-007-0023-7
7. Boston Consulting Group. (2025). Leading in the age of AI agents: Managing the machines that manage themselves. BCG.
8. Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J. D., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., Agarwal, S., Herbert-Voss, A., Krueger, G., Henighan, T., Child, R., Ramesh, A., Ziegler, D., Wu, J., Winter, C., ... Amodei, D. (2020). Language models are few-shot learners. Advances in Neural Information Processing Systems, 33, 1877–1901.
9. Cîrțînă, L.-M., Dumitrașcu, A.-E., Cazacu, D. V., & Ianaşi, C. A. (2025). Eight-disciplines analysis method and quality planning for optimizing problem-solving in the automotive sector: A case study. Processes, 13(10), Article 3121. https://doi.org/10.3390/pr13103121
10. Collins, J. (2001). Good to great: Why some companies make the leap...and others don't. HarperBusiness.
11. Devlin, J., Chang, M. W., Lee, K., & Toutanova, K. (2018). BERT: Pre-training of deep bidirectional transformers for language understanding. arXiv. https://doi.org/10.48550/arXiv.1810.04805
12. Edge, D., Trinh, H., Cheng, N., Bradley, J., Chao, A., Mody, A., Truitt, S., & Larson, J. (2024). From local to global: A graph RAG approach to query-focused summarization. arXiv. https://doi.org/10.48550/arXiv.2404.16130
13. Fragiadakis, G., Tsafarakis, S., & Zopounidis, C. (2024). Evaluating human-AI collaboration: A review and methodological framework. Computers & Industrial Engineering, 189, Article 109962.
14. Gao, Q., Liang, Y., & Zhang, J. (2025). Human-Centered Human-AI Collaboration (HCHAC): Framework and design guidelines. International Journal of Human-Computer Studies, 181, Article 103150.
15. Gao, Y., Xiong, Y., Gao, X., Jia, K., Pan, J., Bi, Y., Dai, Y., Sun, J., Wang, M., & Wang, H. (2024). Retrieval-augmented generation for large language models: A survey. arXiv. https://doi.org/10.48550/arXiv.2312.10997
16. Heredia Álvaro, J. A., & González Barreda, J. (2025). An advanced retrieval-augmented generation system for manufacturing quality control. Advanced Engineering Informatics, 64, Article 102800. https://doi.org/10.1016/j.aei.2024.102800
17. Hevner, A. R. (2007). A three cycle view of design science research. Scandinavian Journal of Information Systems, 19(2), 87–92.
18. Hevner, A. R., March, S. T., Park, J., & Ram, S. (2004). Design science in information systems research. MIS Quarterly, 28(1), 75–105. https://doi.org/10.2307/25148625
19. Huszák, C., Pinke, P., & Kovács, T. A. (2025). Tools for a root cause analysis for safety-critical components: A review. In A. T. Kovács (Ed.), Advanced sciences and technologies for security applications (ICCECIP 2024) (pp. 115–130). Springer.
20. International Automotive Task Force. (2016). IATF 16949:2016 — Quality management system requirements for automotive production and relevant service parts organizations. IATF.
21. Ji, Z., Lee, N., Frieske, R., Yu, T., Su, D., Xu, Y., Ishii, E., Ye, Y. J., Liu, S., & Fung, P. (2023). Survey of hallucination in natural language generation. ACM Computing Surveys, 55(12), Article 248. https://doi.org/10.1145/3571730
22. Kaplík, P., Prístavka, M., Bujna, M., & Viderňan, J. (2013). Use of 8D method to solve problems. Research in Agricultural Engineering, 59(Special Issue), S33–S39.
23. Kempel, M., Richter, R., Deuse, J., Schmid, S., & Schulte, L. (2023). Knowledge graph-based approach for interactive problem solving with the 8D method. Proceedings of the 2023 Smart Systems Integration Conference and Exhibition (SSI), 1–8.
24. Kiangala, K. S., & Wang, Z. (2025). A generative pre-trained transformer industrial bot to improve operators' working experience in a small Industry 5.0 factory. The International Journal of Advanced Manufacturing Technology, 136(3), 1215–1232. https://doi.org/10.1007/s00170-024-14880-w
25. Kumar, S., Verma, M., & Dubey, D. (2023). Reducing the defects and improving the quality of manufacturing product using 8D problem solving tool. Indian Journal of Science and Research, 3(2), 47–55.
26. LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436–444. https://doi.org/10.1038/nature14539
27. Lee, D., Lee, D., & Na, J. (2022). Automatic failure modes and effects analysis of an electronic fuel injection model. Applied Sciences, 12(6), Article 6144. https://doi.org/10.3390/app12066144
28. Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., Küttler, H., Lewis, M., Yih, W., Rocktäschel, T., Riedel, S., & Kiela, D. (2020). Retrieval-augmented generation for knowledge-intensive NLP tasks. Advances in Neural Information Processing Systems, 33, 9459–9474.
29. Li, J., Liu, X., & Zhang, Y. (2023). Applications of large language models in smart manufacturing: A review. Journal of Manufacturing Systems, 68, 120–135. https://doi.org/10.1016/j.jmsy.2023.03.005
30. Liu, Y., & Wang, H. (2026). Who on earth is using Generative AI? World Development, 199, Article 107260. https://doi.org/10.1016/j.worlddev.2025.107260
31. Mahmood, K. (2023). Solving manufacturing problems with 8D methodology: A case study of leakage current in a production company. Journal of Electrical Electronics Engineering, 2(1), 1–18.
32. March, S. T., & Smith, G. F. (1995). Design and natural science research on information technology. Decision Support Systems, 15(4), 251–266. https://doi.org/10.1016/0167-9236(94)00041-2
33. McKinsey & Company. (2025, November). The state of AI in 2025: Agents, innovation, and transformation. McKinsey & Company.
34. Nonaka, I. (1991). The knowledge-creating company. Harvard Business Review, 69(6), 96–104.
35. Nonaka, I. (1994). A dynamic theory of organizational knowledge creation. Organization Science, 5(1), 14–37. https://doi.org/10.1287/orsc.5.1.14
36. Nonaka, I., & Takeuchi, H. (1995). The knowledge-creating company: How Japanese companies create the dynamics of innovation. Oxford University Press.
37. Oancea, A.-V., Cîrțînă, L.-M., Dumitrașcu, A.-E., & Cazacu, D. V. (2025). Integrating Industry 4.0 technologies into 8D methodologies: A case study in the automotive industry. Applied Sciences, 15(20), Article 11262.
38. OpenAI. (2023). GPT-4 technical report. arXiv. https://doi.org/10.48550/arXiv.2303.08774
39. Ouyang, L., Wu, J., Jiang, X., Almeida, D., Wainwright, C., Mishkin, P., Zhang, C., Agarwal, S., Slama, K., Ray, A., Schulman, J., Hilton, J., Kelton, F., Miller, L., Simens, M., Askell, A., Welinder, P., Christiano, P. F., Leike, J., & Lowe, R. (2022). Training language models to follow instructions with human feedback. Advances in Neural Information Processing Systems, 35, 27730–27744.
40. Pan, S., Luo, L., Wang, Y., Chen, C., Wang, J., & Wu, X. (2024). Unifying large language models and knowledge graphs: A roadmap. IEEE Transactions on Knowledge and Data Engineering, 36(7), 3580–3599. https://doi.org/10.1109/TKDE.2023.3321528
41. Peffers, K., Tuunanen, T., Rothenberger, M. A., & Chatterjee, S. (2007). A design science research methodology for information systems research. Journal of Management Information Systems, 24(3), 45–77. https://doi.org/10.2753/MIS0742-1222240302
42. Phanden, R. K., Sheokand, A., Goyal, K. K., Gahlot, P., & Demir, H. I. (2022). 8Ds method of problem solving within automotive industry: Tools used and comparison with DMAIC. Materials Today: Proceedings, 65, 3266–3272. https://doi.org/10.1016/j.matpr.2022.05.419
43. Pietsch, D., Matthes, M., Wieland, U., Ihlenfeldt, S., & Munkelt, T. (2024). Root cause analysis in industrial manufacturing: A scoping review of current research, challenges and the promises of AI-driven approaches. Journal of Manufacturing and Materials Processing, 8(6), Article 277. https://doi.org/10.3390/jmmp8060277
44. Rafailov, R., Sharma, A., Mitchell, E., Ermon, S., Manning, C. D., & Finn, C. (2024). Direct preference optimization: Your language model is secretly a reward model. Advances in Neural Information Processing Systems, 36, 53728–53741.
45. Riesenberger, C. A., & Sousa, S. D. (2010). The 8D methodology: An effective way to reduce recurrence of customer complaints? Proceedings of the World Congress on Engineering 2010, 3, 2225–2230.
46. Rousset, T., Djeddal, L., & Boughanem, M. (2025). Merging language and domain specific models. Information Processing & Management, 62(2), Article 103950.
47. Rumelhart, D. E., Hinton, G. E., & Williams, R. J. (1986). Learning representations by back-propagating errors. Nature, 323(6088), 533–536. https://doi.org/10.1038/323533a0
48. Rusman, K., & Hetharia, D. (2025). Quality improvement through 8D methodology: An automotive industry case study. Operations Excellence: Journal of Applied Industrial Engineering, 17(1), 1–11. https://doi.org/10.22441/oe.2025.v17i1.001
49. Russell, S., & Norvig, P. (2020). Artificial intelligence: A modern approach (4th ed.). Pearson.
50. Scharmer, F., Kaynak, H. (2003). The relationship between total quality management practices and their effects on firm performance. Journal of Operations Management, 21(4), 405–435. https://doi.org/10.1016/S0272-6963(03)00004-4
51. Schiffauerova, A., & Thomson, V. (2006). A review of research on cost of quality models and best practices. International Journal of Quality & Reliability Management, 23(6), 647–669. https://doi.org/10.1108/02656710610672470
52. Schneider, J. (2025). Generative to agentic AI: Survey, conceptualization, and challenges. arXiv. https://doi.org/10.48550/arXiv.2504.18875
53. Shen, B.-M., & Chen, G. Y.-H. (2025). Implementation of Ford 8D in improving efficiencies of wafer testers: A case study. Engineering Management Journal, 37(1), 45–58.
54. Sousa, R., & Voss, C. A. (2002). Quality management re-visited: A reflective review and agenda for future research. Journal of Operations Management, 20(1), 91–109. https://doi.org/10.1016/S0272-6963(01)00088-2
55. Stamatis, D. H. (2003). Failure mode and effect analysis: FMEA from theory to execution (2nd ed.). ASQ Quality Press.
56. Subriadi, A. P., & Najwa, N. F. (2020). The consistency analysis of failure mode and effect analysis (FMEA) in information technology risk assessment. Heliyon, 6(1), Article e03161. https://doi.org/10.1016/j.heliyon.2020.e03161
57. Tao, F., Zhang, M., & Liu, A. (2024). Industrial data flywheel: Driven smart quality control and maintenance in Industry 4.0/5.0. IEEE Transactions on Industrial Informatics, 20(3), 4102–4115. https://doi.org/10.1109/TII.2023.3312520
58. Turing, A. M. (1950). Computing machinery and intelligence. Mind, 59(236), 433–460. https://doi.org/10.1093/mind/LIX.236.433
59. Uslu Divanoğlu, S., & Taş, Ü. (2022). Application of 8D methodology: An approach to reduce failures in automotive industry. Engineering Failure Analysis, 134, Article 106019. https://doi.org/10.1016/j.engfailanal.2021.106019
60. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention is all you need. Advances in Neural Information Processing Systems, 30, 5998–6008.
61. Vial, G. (2019). Understanding digital transformation: A review and a research agenda. The Journal of Strategic Information Systems, 28(2), 118–144. https://doi.org/10.1016/j.jsis.2019.01.003
62. Wu, Z., Liu, W., & Nie, W. (2021). Literature review and prospect of the development and application of FMEA in manufacturing industry. The International Journal of Advanced Manufacturing Technology, 112(5), 1409–1436. https://doi.org/10.1007/s00170-020-06425-0