本研究旨在開發一套基於多代理架構的智慧代理人系統，以業務代理人為例，探討其在醫療面板產業中應對多模態知識檢索與回應挑戰的可行性。醫療面板產業的業務部門需要整合多元資料來源（如產品手冊、報價規範、操作圖表等），並即時提供準確回應。然而，傳統FAQ或關鍵字檢索系統在處理跨來源、多模態資料時，往往面臨知識斷層、回應延遲與正確率不足的問題，進而影響工作效率與客戶滿意度。
為回應此需求，本研究提出一個多代理人架構，包含主控代理人（Master Agent）及三個子代理人（知識型、規則型與分析型），各子代理人分工處理不同知識任務，並由主控代理人協調資源與任務分配，以達到最佳化效能。
技術創新主要體現在混合檢索策略的應用。系統結合知識圖譜與向量檢索技術，實現跨模態資料的整合檢索，彌補傳統檢索無法有效處理多樣化資料的不足，進而提升回應的即時性與準確性。此外，透過引入GraphRAG技術，系統在資料檢索的廣度與深度上更為全面。另一方面，本研究設計了Routing策略，能依據任務特性動態分配代理人角色，確保每筆查詢獲得最合適的處理，從而滿足不同業務需求並提升整體決策支援能力。

This study proposes a multi-agent intelligent system designed to enhance multi-modal knowledge retrieval and response in the business departments of the medical panel industry. Traditional keyword-based systems often struggle with integrating diverse data formats—such as manuals, pricing rules, and diagrams—leading to delayed and inaccurate responses. To address this, the system adopts a modular architecture comprising a Master Agent and three specialized sub-agents (knowledge-based, rule-based, and analysis-based), each responsible for handling different task types under dynamic coordination.
The system leverages a hybrid retrieval approach that combines knowledge graphs and vector search, enhanced by GraphRAG, to improve the accuracy and relevance of responses across modalities. A flexible routing mechanism further ensures that queries are assigned to the most appropriate agent, improving efficiency and adaptability.
The proposed architecture not only improves business response capabilities but also offers a scalable framework for intelligent agents in broader enterprise applications. Future work will focus on completing the remaining agent modules and expanding real-world deployment scenarios.

Acloudear. (2024). Inquiry and procurement process in manufacturing enterprises. Acloudear.
Anthropic. (2025, June 13). How we built our multi-agent research system. Anthropic Blog.
Anonymous. (2025). Review of autonomous and collaborative agentic AI and multi-agent systems for enterprise applications. International Journal of Innovative Research in Engineering & Management (IJIREM).
Brynjolfsson, E., Li, D., & Raymond, L. R. (2023). Generative AI at work. National Bureau of Economic Research Working Paper.
Cai, F., Zhou, W., Mi, F., & Faltings, B. (2021). SLIM: Explicit slot-intent mapping with BERT for joint multi-intent detection and slot filling. arXiv preprint arXiv:2105.01929.
Chen, S., Liu, Y., Han, W., Zhang, W., & Liu, T. (2024). A survey on LLM-based multi-agent system: Recent advances and new frontiers in application. arXiv preprint arXiv:2412.17481.
Chen, W., Hu, H., Chen, X., Verga, P., & Cohen, W. (2022). MuRAG: Multimodal retrieval-augmented generator for open question answering over images and text. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing (pp. 5558–5570). Association for Computational Linguistics.
Ding, Y., Facciani, M., Poudel, A., Joyce, E., Aguinaga, S., Veeramani, B., Bhattacharya, S., & Weninger, T. (2025). Citations and trust in LLM generated responses. In Proceedings of the AAAI Conference on Artificial Intelligence (Vol. 39, No. 22, pp. 23787–23795).
Edge, D., Larson, J., Truitt, S., Mody, A., Chao, A., Bradley, J., … (2024). From local to global: A Graph RAG approach to query-focused summarization. ICLR submission. OpenReview.
EICopilot. (2025). EICopilot: Search and explore enterprise information over Graph2NL and ICL. arXiv preprint arXiv:2501.13746.
Gao, Y., Xiong, Y., Gao, X., Jia, K., Pan, J., Bi, Y., Dai, Y., Sun, J., Guo, Q., Wang, M., & Wang, H. (2023). Retrieval-augmented generation for large language models: A survey. arXiv preprint arXiv:2312.10997.
Gelashvili-Luik, T., Vihma, P., & Pappel, I. (2025). Navigating the AI revolution: Challenges and opportunities for integrating emerging technologies into knowledge management systems—Systematic literature review. Frontiers in Artificial Intelligence, 8, 1595930.
Gupta, S., Ranjan, R., & Singh, S. N. (2024). A comprehensive survey of retrieval-augmented generation (RAG): Evolution, current landscape and future directions. arXiv preprint arXiv:2410.12837.
Han, H., Wang, Y., Shomer, H., Guo, K., Ding, J., Lei, Y., … Tang, J. (2025). Graph-R1: Towards agentic GraphRAG framework via end-to-end reinforcement learning. arXiv preprint.
Hogan, A., Blomqvist, E., Cochez, M., d'Amato, C., de Melo, G., Gutiérrez, C., … Kirrane, S. (2021). Knowledge graphs. ACM Computing Surveys, 54(4), 71.
Hu, Y., Lei, Z., Zhang, Z., Pan, B., Ling, C., & Zhao, L. (2024). GRAG: Graph retrieval-augmented generation. arXiv preprint.
Krishnan, N. (2025). Advancing multi-agent systems through model context protocol. arXiv preprint arXiv:2504.21030.
Ledro, C., Nosella, A., Vinelli, A., Dalla Pozza, I., & Souverain, T. (2025). Artificial intelligence in customer relationship management: A systematic framework for a successful integration. Journal of Business Research, 199, 115531.
Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., … & Riedel, S. (2020). Retrieval-augmented generation for knowledge-intensive NLP tasks. arXiv preprint arXiv:2005.11401.
Li, X., Wang, S., Zeng, S., Wu, Y., & Yang, Y. (2024). A survey on LLM-based multi-agent systems: Workflow, infrastructure, and challenges. Computational Visual Media. Springer Nature.
Liu, J., Tan, Y. K., Fu, B., & Lim, K. H. (2024). Intent-aware dialogue generation and multi-task contrastive learning for multi-turn intent classification. arXiv preprint.
Liu, P., Liu, X., Yao, R., Liu, J., Meng, S., Wang, D., & Ma, J. (2025). Hierarchical multi-agent multimodal retrieval-augmented generation (HM-RAG). arXiv preprint.
MarketsandMarkets. (2023). Multimodal AI market by offering, modality, technology, application, and region—Global forecast to 2028. MarketsandMarkets.
Martis, L. (2024, December 10). LuminiRAG: Vision-enhanced Graph RAG for complex multi-modal document understanding. TechRxiv.
McKinsey & Company. (2023). The state of AI in 2023: Generative AI’s breakout year. McKinsey Global Institute.
Nascimento, N., Alencar, P., & Cowan, D. (2023). Self-adaptive large language model (LLM)-based multiagent systems. arXiv preprint arXiv:2307.06187.
Neo4j Blog. (2024, June 5). Knowledge graph vs. vector RAG: Optimization & analysis. Neo4j.
NVIDIA Developer Blog. (2024, March 20). An easy introduction to multimodal retrieval-augmented generation. NVIDIA.
Plekhanov, D., Franke, H., & Netland, T. H. (2023). Digital transformation: A review and research agenda. European Management Journal, 41(6), 821–844.
Radford, A., Kim, J. W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., … & Sutskever, I. (2021). Learning transferable visual models from natural language supervision. In Proceedings of the 38th International Conference on Machine Learning (ICML) (Vol. 139, pp. 8748–8763). PMLR.
Rožanec, J. M., Zajec, P., Kenda, K., Novalija, I., Fortuna, B., & Mladenić, D. (2021). XAI-KG: Knowledge graph to support XAI and decision-making in manufacturing. arXiv preprint arXiv:2105.01929.
Schneider, P., Schopf, T., Vladika, J., & Matthes, F. (2024). Enterprise use cases combining knowledge graphs and natural language processing. In Informing Possible Future Worlds: Essays in Honour of Ulrich Frank (pp. 271–285). Logos Verlag Berlin GmbH.
Snowflake Blog. (2025, April 21). Evaluating multimodal vs. text-based retrieval for RAG with Snowflake Cortex. Snowflake.
Superteams.ai Blog. (2024, November 30). A deep-dive into vector search and knowledge graph for AI. Superteams.ai.
Tran, K.-T., Dao, D., Nguyen, M.-D., Pham, Q.-V., O’Sullivan, B., & Nguyen, H. D. (2025). Multi-agent collaboration mechanisms: A survey of LLMs. arXiv preprint arXiv:2501.06322.
Verhoef, P. C., Broekhuizen, T., Bart, Y., Bhattacharya, A., Dong, J. Q., Fabian, N., & Haenlein, M. (2021). Digital transformation: A multidisciplinary reflection and research agenda. Journal of Business Research, 122, 889–901.
Wang, Z., Panda, R., Karlinsky, L., Feris, R., Sun, H., & Kim, Y. (2023). Multitask prompt tuning enables parameter-efficient transfer learning. arXiv preprint.
Wikipedia. (2025). Multimodal representation learning. In Wikipedia.
Wu, F., Li, Z., Wei, F., Li, Y., Ding, B., & Gao, J. (2025). Talk to right specialists: Routing and planning in multi-agent system for question answering. arXiv preprint.
Yan, B., Zhang, X., Zhang, L., Zhang, L., Zhou, Z., Miao, D., & Li, C. (2025). Beyond self-talk: A communication-centric survey of LLM-based multi-agent systems. arXiv preprint arXiv:2502.14321.
Yue, Y., Zhang, G., Liu, B., Wan, G., Wang, K., Cheng, D., & Qi, Y. (2025). MasRouter: Learning to route LLMs for multi-agent systems. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Long Papers).
Zhu, X., Xie, Y., Liu, Y., Li, Y., & Hu, W. (2025). Knowledge graph-guided retrieval augmented generation (KG²RAG). In Proceedings of NAACL 2025 Long Papers (pp. 8912–8924). Association for Computational Linguistics.