本研究聚焦於當代高科技工業設施建造時，面臨海量且繁瑣之設計法規的知識管理痛點，指出傳統依賴專家人力審查的模式已遭逢人才斷層的瓶頸。為解決此困境，本研究結合大型語言模型（Large Language Model, LLM)微調技術，開發一套能自動閱讀、精準分類並結構化輸出的輔查系統。以 M 集團專案為實證對象，將專家的隱性經驗轉化為外顯知識圖譜，為工程業界提供數位轉型方案與建立動態能力的基礎。在前處理階段，實施機密資訊遮罩、視覺語言模型文本重建，並採用 7:3 比例的真實與生成樣本擴增策略以克服企業資料稀缺的限制。在核心分類上，選用 LLaMA 3.1 8B 模型進行階層式句子與實體分類，並將結構化數據匯入圖資料庫建構知識圖譜。最終將技術整合至圖形化介面，實現自動產出檢核表與關聯拓撲圖。
實作結果顯示，樣本擴增(Data Augmentation)策略大幅降低了企業導入 AI 的門檻。針對實體細部子領域分類準確率較低的問題，本研究提出動態提示工程、雙重檢核機制與未分類（Unclassified）容錯設計等優化策略。為客觀評估系統導入成效，本研究採量化績效指標與質化訪談編碼之雙軌評估設計，實務測試證實，系統能極大化法規處理與跨部門分派的效率，並透過知識圖譜視覺化有效打破資訊孤島，充分展現了組織在專案管理中感知與掌握合規風險的動態能力。實務上，系統透過人機協作機制促成審查決策前移，並將經驗教訓（Lessons Learned）持續反饋至模型中，落實企業內部隱性知識的外顯化與動態擴增來達成有效的知識管理。未來研究可進一步擴展跨行業規範比較與指令複雜度的系統性實驗，確保企業在快速變動環境下，能透過數位轉型與動態能力的實踐維持長期的競爭優勢。

Engineering projects such as semiconductor manufacturing require precise compliance with vast, evolving design codes, yet traditional review remains dependent on scarce senior engineering talent-as exemplified by M Group's expansion into semiconductor manufacturing. This study proposes a hybrid system, the Augmented Engineering Review (AER) framework, integrating large language models (LLMs) with LoRA fine-tuning and knowledge graphs to automatically classify regulatory text, generate checklists, and visualize cross-domain code relationships.
Grounded in Dynamic Capabilities Theory (Teece, 2007), the system's synonym-expansion, LLM classification, and human-in-the-loop mechanisms map onto sensing, seizing, and transforming capabilities. Trained on a 70:30 ratio of authentic-to-synthetic data on on-premises hardware, the classification pipeline evolved from word-level BERT to sentence-level MPNet to a final LLaMA 3.1 8B configuration achieving 87.47% accuracy on large-domain classification and 70% on subdomains, versus 68%/42% for an unfine-tuned baseline.
A/B testing showed processing time falling from 1-4 hours to about 3 minutes per document and decision confidence rising from 3.2 to 4.6. Qualitative findings indicated reduced inter-departmental ambiguity, conflicts surfaced via knowledge-graph visualization, and increased trust via an “Unclassified” fail-safe mechanism.
These findings suggest competitive advantage may depend more on data representativeness than volume, and that AI can transform workflows by relocating neglected decisions earlier in the process. The approach lowers the resource threshold for AI adoption among small and mid-sized firms, though limitations include single-enterprise data and limited sample sizes, pointing toward cross-industry and longitudinal study.

Barney, J. (1991). Firm resources and sustained competitive advantage. Journal of management, 17(1), 99-120.
de Magalhães Santos, L. G. (2024). Dynamic capabilities in the public sector to deal with GovTech. International Conference on Electronic Government,
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2019). Bert: Pre-training of deep bidirectional transformers for language understanding. Proceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologies, volume 1 (long and short papers),
Eastman, C., Lee, J.-m., Jeong, Y.-s., & Lee, J.-k. (2009). Automatic rule-based checking of building designs. Automation in construction, 18(8), 1011-1033.
Fitkau, I., & Hartmann, T. (2024). An ontology-based approach of automatic compliance checking for structural fire safety requirements. Advanced Engineering Informatics, 59, 102314.
Hassani, S., Sabetzadeh, M., Amyot, D., & Liao, J. (2024). Rethinking legal compliance automation: Opportunities with large language models. 2024 IEEE 32nd International Requirements Engineering Conference (RE),
Hu, E. J., Shen, Y., Wallis, P., Allen-Zhu, Z., Li, Y., Wang, S., Wang, L., & Chen, W. (2022). Lora: Low-rank adaptation of large language models. Iclr, 1(2), 3.
Iversen, O., & Huang, L. (2026). Leveraging large language models for BIM-based automated compliance checking. Automation in construction, 182, 106707.
Jiang, L., Shi, J., & Wang, C. (2022). Multi-ontology fusion and rule development to facilitate automated code compliance checking using BIM and rule-based reasoning. Advanced Engineering Informatics, 51, 101449.
Kraus, S., Jones, P., Kailer, N., Weinmann, A., Chaparro-Banegas, N., & Roig-Tierno, N. (2021). Digital transformation: An overview of the current state of the art of research. Sage Open, 11(3), 21582440211047576.
Li, J., & Maiti, A. (2025). Applying large language model analysis and backend web services in regulatory technologies for continuous compliance checks. Future Internet, 17(3), 100.
Nonaka, L., Takeuchi, H., & Umemoto, K. (1996). A theory of organizational knowledge creation. International journal of technology Management, 11(7-8), 833-845.
PMI. (2021). A guide to the project management body of knowledge (7 ed.). Project Management Institute.
Purushotham, N., Kailashnath, C., & Mutis, I. (2026). Framework for automated building code compliance checking to improve transparency, trust, validation, and design interpretation. Automation in construction, 181, 106598.
Purushotham, P., & Jagadish, N. (2026). Mathematical framework for measuring and enhancing cybersecurity awareness against phishing attacks. 2026 International Conference on Smart Futuristic Technology,
Teece, D. J. (2007). Explicating dynamic capabilities: The nature and microfoundations of (sustainable) enterprise performance. Strategic management journal, 28(13), 1319-1350.
Teece, D. J., Pisano, G., & Shuen, A. (1997). Dynamic capabilities and strategic management. Strategic management journal, 18(7), 509-533.
Xu, Z., Sheng, Y., Bao, Q., Du, X., Guo, X., & Liu, Z. (2025). BERT-based automatic audit report generation and compliance analysis. 2025 5th International Conference on Artificial Intelligence and Industrial Technology Applications (AIITA),
Xue, X., & Zhang, J. (2022). Regulatory information transformation ruleset expansion to support automated building code compliance checking. Automation in construction, 138, 104230.
Zu, F., & Zhang, X. (2026). Real-time knowledge management for construction value engineering: Live capture and BERT-aided case-based retrieval. Automation in construction, 183, 106782.