跳到主要內容

簡易檢索 / 詳目顯示

回結果列表
研究生： 劉冠吟
Liu, Kuan-Yin
論文名稱： 整合 Agentic AI 要素於金融 AI 雲平台之研究
Integrating the Agentic AI elements into the Financial AI Cloud
指導教授： 蔡瑞煌
Tsaih, Rua-Huan
張欣綠
Chang, Hsin-Lu
口試委員: 吳文舜
Wu, Wen-Shuen
學位類別： 碩士
Master
系所名稱： 商學院 - 資訊管理學系
Department of Management Information System
論文出版年： 2026
畢業學年度： 114
語文別： 英文
論文頁數： 114
中文關鍵詞： 代理式AI 、AI代理人 、金融AI雲 、FinAIC 、AI技術堆疊 、多代理人系統 、代理人治理 、主權AI 、金融科技
外文關鍵詞： Agentic AI, AI Agent, Financial AI Cloud, FinAIC, AI tech-stack, Multi-agent system, Agent governance, Sovereign AI, FinTech
相關次數： 點閱：6  下載：0 
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報

    • 隨著生成式人工智慧、大型語言模型與 AI Agent 的快速發展，金融 AI 系統正由傳統辨識型 AI 與 API 式服務，轉向具備語意理解、任務規劃、工具調用、記憶管理與多代理人協作能力的 Agentic AI 系統。然而，金融機構在導入 Agentic AI 時，仍面臨資料治理、系統整合、資安控管、法規責任、算力成本與組織成熟度等挑戰，使既有金融 AI 雲平台架構有重新檢視與延伸之必要。

    本研究以金融 AI 雲平台（Financial AI Cloud, FinAIC）為基礎，探討如何將 Agentic AI 要素整合至既有七層技術堆疊中。研究首先藉由相關文獻參考，並思考 Agentic AI 時代下 AI 系統所需具備的關鍵要素，將 AI Tech-Stack 延伸為 Agentic AI Tech-Stack；其次，進一步定義 Agentic AI Cloud 的一般性架構，說明 Model、Orchestration、Tools、AgentOps、多代理人協作與治理機制如何嵌入雲端技術堆疊；最後，將此架構實例化至金融場域，提出 Agentic FinAIC 架構，使 FinAIC 由傳統資源與服務供應平台，轉型為任務導向、模組化且可治理的金融 Agentic AI 平台。本研究並透過三位金融科技與金融資訊領域專家之深度訪談，驗證此架構於台灣金融產業導入之可行性與必要性。

    研究結果顯示，台灣金融業導入 Agentic AI 仍處於探索與內部試點階段，較適合優先應用於內部知識整合、法遵檢查、風險預警、資料整理、SOP 查詢、投資研究輔助與人機協作場景，而不宜立即導入高風險、對外且全自動之金融決策。其次，Agentic FinAIC 的核心價值不在於單一模型或算力資源，而在於以一致標準整合資料、API、模型、工具、代理人協作與治理機制。金融 Agentic AI 應採取分層共享與混合式部署模式，並以 bounded autonomy 限定代理人之自主邊界、授權範圍、操作紀錄與責任歸屬。

    本研究之貢獻在於延伸 AI Tech-Stack 與 FinAIC 架構，提出具備模組化 AI Agent、多代理人協作、平台治理與主權 AI 意涵的 Agentic FinAIC 架構，可作為金融機構、技術供應商、監理機關與產業公會評估金融 Agentic AI 導入策略與治理模式之參考。

    With the rapid development of Generative AI, Large Language Models, and AI agents, financial AI systems are shifting from traditional discriminative AI and API-based services toward Agentic AI systems capable of semantic understanding, task planning, tool invocation, memory management, and multi-agent collaboration. However, financial institutions still face challenges in data governance, system integration, cybersecurity, regulatory accountability, computing cost, and organizational readiness. These challenges indicate the need to revisit and extend the existing architecture of financial AI cloud platforms.

    This study builds on the Financial AI Cloud (FinAIC) and explores how Agentic AI elements can be integrated into its seven-layer technology stack. Based on literature references and conceptual reasoning on the key requirements of AI systems in the agentic era, this study first extends the AI Tech-Stack into the Agentic AI Tech-Stack. It then further defines the general architecture of an Agentic AI Cloud by explaining how Model, Orchestration, Tools, AgentOps, multi-agent collaboration, and governance mechanisms can be embedded into a cloud-based technology stack. Finally, this architecture is instantiated in the financial domain as Agentic FinAIC, transforming FinAIC from a resource and service provision platform into a task-oriented, modular, and governable financial Agentic AI platform. Three expert interviews with professionals in FinTech and financial information management are conducted to examine the feasibility and necessity of the proposed architecture in Taiwan’s financial industry.

    This study contributes by extending the AI Tech-Stack and FinAIC architectures and proposing Agentic FinAIC as a modular, multi-agent, governance-oriented, and sovereign AI-oriented financial AI platform. The findings provide a reference for financial institutions, technology providers, regulators, and industry associations in evaluating Agentic AI adoption strategies and governance models.

    摘要 ii
    Abstract iii
    Table of Contents v
    Tables viii
    Figures ix
    Chapter 1 Introduction 1
    1.1 Research Background 1
    1.2 Research Objectives 2
    Chapter 2 Literature Review 5
    2.1 Evolution and Key Technologies of LLMs 5
    2.1.1 Development and Trends of Large Language Models 5
    2.1.2 Domain Adaptation and Performance Optimization Strategies 5
    2.2 AI Agent Cognitive Architecture and Workflows 6
    2.2.1 Definition of AI Agents and the Trinity Framework 6
    2.2.2 Orchestration Patterns and Reasoning Loops 7
    2.2.3 Information Retrieval Evolution: From RAG to Agentic RAG 8
    2.2.4 Multi-Agent Systems (MAS) and Coordination 8
    2.3 Financial AI Cloud Platforms, AI Tech Stack and FinAIC 8
    2.3.1 AI Tech-Stack Model 8
    2.3.2 Financial AI Cloud 9
    2.3.3 The Seven-Layer Stack Architecture of FinAIC 10
    2.3.4 Limitations of Original Cloud Architectures in the Agentic Era 13
    2.3.5 Research Gap and Synthesis 14
    Chapter 3 Proposed Agentic FinAIC Architecture Design 16
    3.1 From AI Tech-Stack to Agentic AI Tech-Stack 16
    3.1.1 Agentic AI Elements Added to the AI Tech-Stack 17
    3.1.2 The Proposed Agentic AI Tech-Stack 20
    3.2 Agentic AI Cloud: A General Cloud Architecture 23
    3.2.1 Modular Design of Agentic AI Agents 24
    3.2.2 Multi-Agent System and Agent Collaboration 29
    3.2.3 Agentic AI Tech-Stack Layers and Cloud Function Blocks 31
    3.3 Implementation Issues for a Sovereign Agentic AI Cloud 35
    3.3.1 Security and Compliance 36
    3.3.2 Performance and Latency 37
    3.3.3 Scalability 39
    3.3.4 Collaboration and Reproducibility 40
    3.3.5 Cost-effectiveness and Sustainable Operation 41
    3.3.6 Summary 43
    3.4 Agentic FinAIC as a Financial Instantiation of Sovereign Agentic AI Cloud 43
    3.4.1 Positioning Agentic FinAIC 44
    3.4.2 Seven-Layer Architecture of Agentic FinAIC 47
    3.4.3 Summary and Transition to Expert Evaluation 50
    Chapter 4 Expert In-depth Interviews 52
    4.1 Interview Design and Analytical Approach 52
    4.2 Role-based Interview Summaries 56
    4.2.1 FinTech and Digital Transformation Expert Perspective 56
    4.2.2 Banking Digital Finance Executive Perspective 59
    4.2.3 Financial Holding IT and Cybersecurity Executive Perspective 63
    4.3 Thematic Analysis and Cross-role Comparison 69
    4.3.1 Early Demand for Agentic FinAIC 69
    4.3.2 Modular and Shared Adoption Mechanism 69
    4.3.3 Data Pipeline, Orchestration, and API Integration as Core Requirements 70
    4.3.4 Agent Governance and Risk-based Automation 71
    4.3.5 Ecosystem Participation and Differentiated Value 71
    4.3.6 Summary of Cross-role Findings 72
    Chapter 5 Research Findings and Architectural Reflection 77
    5.1 Research Findings and Reflection on the Proposed Architecture 77
    5.1.1 Agentic AI Changes the Service Logic of FinAIC from Resource Provision to Task-oriented Execution 77
    5.1.2 The Key Architectural Value of Agentic FinAIC Lies in Governed Integration 79
    5.1.3 Bounded Autonomy Should Be Considered at the Institution-wise, Industry Management, and Regulatory Governance Levels 81
    5.1.4 Agent Governance Requires Operationalization through Domain Management, Bounded Autonomy, and Business Logic Management 83
    5.2 Business Logic as a Cross-layer Governance Concern in Agentic FinAIC 86
    5.2.1 From Implicit Application Logic to Explicit Architectural Issue 86
    5.2.2 Scope of Cross-layer Business Logic Governance 87
    5.2.3 Proposed Business Logic Content 89
    5.2.4 Further Questions: Toward a More Complete Business Logic Management Framework 91
    Chapter 6 Conclusion 94
    6.1 Research Summary 94
    6.2 Research Contributions 95
    6.2.1 Theoretical Contributions 95
    6.2.2 Practical Contributions 96
    6.3 Research Limitations 97
    6.4 Future Research Works 98
    6.5 Conclusion 101
    Reference 102
    Appendix A. Interview Protocol 106
    Appendix B. Thematic Interview Records 109

    Antonio Gulli, Lavi Nigam, Julia Wiesinger, Vladimir Vuskovic, Irina Sigler, Ivan Nardini, Nicolas Stroppa, Sokratis Kartakis, Narek Saribekyan, Anant Nawalgaria, & Alan Bount. (2025). Agents Companion. https://www.kaggle.com/whitepaper-agent-companion
    Banh, L., & Strobel, G. (2023). Generative artificial intelligence. Electronic Markets, 33(1). https://doi.org/10.1007/s12525-023-00680-1
    Blount, A., Gulli, A., Saboo, S., Zimmermann, M., Vuskovic, V., Chan, E., Clark, M., Egan, D., Nawalgaria, A., Patlolla, K., & Wiesinger, J. (2025). Introduction to Agents. https://www.kaggle.com/whitepaper-introduction-to-agents
    Chang, P. Y.-C., Ma, M., & Pflugfelder, B. (2024). Generative AI Agents in Action: Revolutionizing Software Development Testing. https://www.appliedai.de/en/insights/generative-ai-agents-in-action/
    Ding, Q., Ding, D., Wang, Y., Guan, C., & Ding, B. (2024). Unraveling the landscape of large language models: a systematic review and future perspectives. Journal of Electronic Business & Digital Economics, 3(1), 3–19. https://doi.org/10.1108/jebde-08-2023-0015
    Dong, X., Zhang, X., Bu, W., Zhang, D., & Cao, F. (2024). A Survey of LLM-based Agents: Theories, Technologies, Applications and Suggestions. 2024 3rd International Conference on Artificial Intelligence, Internet of Things and Cloud Computing Technology, AIoTC 2024, 407–413. https://doi.org/10.1109/AIoTC63215.2024.10748304
    Gupta, P., Ding, B., Guan, C., & Ding, D. (2024). Generative AI: A systematic review using topic modelling techniques. Data and Information Management, 8(2). https://doi.org/10.1016/j.dim.2024.100066
    Hsu, C.-C. (2022). The AI Tech Stack Model [National Chengchi University]. https://doi.org/https://nccur.lib.nccu.edu.tw/handle/140.119/142643
    Kandragula, S. (2024). Machine Learning and Artificial Intelligence in Cloud Computing. INTERNANTIONAL JOURNAL OF SCIENTIFIC RESEARCH IN ENGINEERING AND MANAGEMENT, 08(10), 1–4. https://doi.org/10.55041/IJSREM17662
    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. In H. Larochelle, M. Ranzato, R. Hadsell, M. F. Balcan, & H. Lin (Eds.), Advances in Neural Information Processing Systems (Vol. 33, pp. 9459–9474). Curran Associates, Inc. https://proceedings.neurips.cc/paper_files/paper/2020/file/6b493230205f780e1bc26945df7481e5-Paper.pdf
    Linkon, A. A., Shaima, M., Sarker, M. S. U., Badruddowza, Nabi, N., Rana, M. N. U., Rahman, S. K. G. M. A., Esa, H., & Chowdhury, F. R. (2024). Advancements and Applications of Generative Artificial Intelligence and Large Language Models on Business Management: A Comprehensive Review. Journal of Computer Science and Technology Studies, 6(1), 225–232. https://doi.org/10.32996/jcsts
    Ma, C., Zhang, J., Li, Z., & Xu, S. (2023). Multi-agent deep reinforcement learning algorithm with trend consistency regularization for portfolio management. Neural Computing and Applications, 35(9), 6589–6601. https://doi.org/10.1007/s00521-022-08011-9
    Mell, P., & Grance, T. (n.d.). The NIST Definition of Cloud Computing Recommendations of the National Institute of Standards and Technology.
    Mienye, I. D., & Swart, T. G. (2024). A Comprehensive Review of Deep Learning: Architectures, Recent Advances, and Applications. Information (Switzerland), 15(12). https://doi.org/10.3390/info15120755
    Panikkar, R., Saleh, T., Szybowski, M., & Whiteman, R. (2021). Operationalizing machine learning in processes. https://www.mckinsey.com/capabilities/operations/our-insights/operationalizing-machine-learning-in-processes
    Patil, R., & Gudivada, V. (2024). A Review of Current Trends, Techniques, and Challenges in Large Language Models (LLMs). In Applied Sciences (Switzerland) (Vol. 14, Number 5). Multidisciplinary Digital Publishing Institute (MDPI). https://doi.org/10.3390/app14052074
    Tsaih, R. H., Chang, H. L., Hsu, C. C., & Yen, D. C. (2023). The AI Tech-Stack Model. Communications of the ACM, 66(3), 69–77. https://doi.org/10.1145/3568026
    Vuletić, M., Felix, P., & and Cucuringu, M. (2024). Fin-GAN: forecasting and classifying financial time series via generative adversarial networks. Quantitative Finance, 24(2), 175–199. https://doi.org/10.1080/14697688.2023.2299466
    Wiesinger, J., Marlow, P., & Vuskovic, V. (2025). Agents. https://www.kaggle.com/whitepaper-agents
    World Economic Forum & Capgemini. (2024). Navigating the AI Frontier: A Primer on the Evolution and Impact of AI Agents. https://www.weforum.org/publications/navigating-the-ai-frontier-a-primer-on-the-evolution-and-impact-of-ai-agents/
    Wu, W.-S. (2024). Financial AI Cloud Take Taiwan’s securities industry as an example. [National Chengchi University]. https://nccur.lib.nccu.edu.tw/handle/140.119/150256
    Yao, S., Zhao, J., Yu, D., Du, N., Shafran, I., Narasimhan, K., & Cao, Y. (2023). ReAct: Synergizing Reasoning and Acting in Language Models. http://arxiv.org/abs/2210.03629

    無法下載圖示 全文公開日期 2031/06/09
    QR CODE
    :::