近年來，隨著數位轉型與生成式人工智慧技術的發展，企業必須持續檢視並調整經營模式，以因應市場需求與競爭環境的變化。然而，現有經營模式分析工具多著重於靜態架構呈現，缺乏對元件間因果關係的視覺化分析，亦未能整合人機協作與知識持續累積機制，難以滿足企業於經營模式分析與決策規劃的實際需求。 為解決上述問題，本研究建構一套 AI 輔助經營模式創新視覺化平台。平台以因果脈絡圖呈現經營模式元件及其關係，並整合大型語言模型 (Large Language Models， LLM )、檢索增強生成 (Retrieval-Augmented Generation， RAG )、人機協作 (Human-in-the-Loop， HITL ) 及知識動態更新機制，使用者可以檢視及修改 AI 建議，確認後的內容則作為知識庫後續更新的資料。 研究採混合研究方法，透過使用者實驗、Likert 量表、操作紀錄分析及半結構式訪談進行初步評估。實驗結果顯示，本平台支援使用者進行經營模式分析與視覺化思考，並有助於促進思考擴展與因果關係建構。在團體情境中，平台亦能協助成員釐清觀點、整合討論內容，並有助於形成共同理解。此外，知識回寫結果初步顯示，平台具有跨使用者累積與再利用分析知識的可能性，但其知識品質與實際支援效果仍需後續研究進一步驗證。本研究成果可作為企業策略分析與團隊決策的輔助參考，並為 AI 輔助經營模式分析、人機協作及知識動態更新機制的設計與應用，提供初步的實證觀察。;As digital transformation and generative artificial intelligence continue to advance， firms must adapt their business models to changing market demands and competitive environments. Existing business model analysis tools mainly present static structures， provide limited support for visualizing causal relationships among business model components， and seldom integrate human–AI collaboration with mechanisms for continuous knowledge accumulation. This study develops an AI-assisted visual platform for business model innovation. The platform uses causal maps to visualize business model components and their relationships while integrating large language models (LLM)， retrieval-augmented generation (RAG)， human-in-the-loop (HITL)， and a dynamic knowledge updating mechanism. Users can inspect and revise AI-generated suggestions， and confirmed content is stored in the knowledge base for future reuse. A mixed-methods approach was adopted. The platform was evaluated using Likert-scale questionnaires， operation logs， and semi-structured interviews. The results indicate that the platform supported business model analysis by helping participants extend ideas and construct causal relationships. In group settings， participants reported that the visual representation facilitated communication， helped organize discussions， and promoted shared understanding. The knowledge write-back records showed overlaps between components contributed by earlier and later participants， suggesting the potential for cross-user knowledge accumulation and reuse， although direct retrieval from the knowledge base was not verified. Further validation through larger samples， source tracking， and expert evaluation is still required. The platform may serve as a decision-support tool for strategic analysis and team discussion. This study provides initial empirical insights into the design of AI-assisted business model analysis， human–AI collaboration， and dynamic knowledge updating mechanisms.
