本研究以採購管理情境為個案，探討企業內部部署的大型語言模型（LLM）在供應商分析、價格談判與競標模擬中的適用性。研究透過多項實驗—供應商健康分析、ZOPA 及 BATNA 建構、以及封標競標（sealed bid）模擬—評價企業自建 GPT 與資深經理人經驗差異，以及與外部開放式 LLM 的推論能力差異。結果顯示企業內部模型能提供一定程度的邏輯推理與協助，但受限於資料即時性與資料封閉不完整、法遵限制與隱私控管，其應用深度仍停留在建議式的「對話式分析」階段，尚未達到做出重大決策建議的高度。
研究亦發現，LLM 在競標模擬中雖能作出合理推論，卻無法完全掌握真實市場中賽局的競價行為，容易低估人類供應商在多重心理壓力下所做出的反應。本研究提出企業要邁向自動化 AI 代理人所需的基礎建設，包括外部資料接入、安全測試沙盒、跨系統整合與監督機制。最後，本研究提出 N 公司內部 GPT 應用的改善方向，以提升採購效率並強化供應鏈競爭力。

This study examines the effectiveness of enterprise-deployed large language models (LLMs) in supporting procurement decision-making within N Company. Through experiments involving supplier analysis, ZOPA/BATNA construction, and sealed-bid simulations, the results show that internal GPT systems can provide basic logical reasoning but remain constrained by closed data environments, limited contextual information, and strict AI-governance requirements. Consequently, their analytical value is restricted to first-layer conversational support rather than strategic decision inference. The study further reveals that LLMs systematically underestimate real-world bidding behavior because critical determinants—such as loss aversion, relational history, internal politics, and tacit knowledge—are inaccessible to the model. Based on these findings, the study outlines the technical and governance prerequisites for advancing toward agentic AI, including controlled external data access, secure sandbox environments, cross-system integration, and human-in-the-loop oversight. Recommendations are provided to guide N Company in enhancing internal GPT adoption and strengthening procurement and supply-chain decision capabilities.

Keyword: Large language models (LLMs), Procurement negotiation, Tacit knowledge, Behavioral decision-making, Supply chain strategy, AI agents, ZOPA, BATNA, Sealed-bid competition, AI governance, Organizational information asymmetry

AI ＆ LLM 類文獻
McKinsey & Company. (2023). The state of AI in 2023: Generative AI’s breakout year. McKinsey Global Survey.
Haenlein, M., & Kaplan, A. (2019). A brief history of artificial intelligence: On the past, present, and future of artificial intelligence. California Management Review, 61(4), 5-14.
Bommasani, R., Hudson, D. A., Adeli, E., Altman, R., Arora, S., von Arx, S., … Liang, P. (2021). On the opportunities and risks of foundation models. Stanford Institute for Human-Centered Artificial Intelligence (HAI).
Abou Ali, M., & Dornaika, F. (2025). Agentic AI: A comprehensive survey of architectures, applications, and future directions. Artificial Intelligence Review, 59, Article 11.
Kabir, S., Udo-Imeh, D. N., Kou, B., & Zhang, T. (2024). Large language models don’t always perform as expected: A study on code-generation tasks. In Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency (FAccT’24) (pp. 1–15). ACM.
Kalai, A. T., Bitton, Y., Edelman, B., Gutman, Y., Novo, S., & Saran, A. (2025). Why language models hallucinate. arXiv:2401.02892. https://arxiv.org/abs/2401.02892
Huang, C.-H., Chang, C.-H., & Lin, Y.-H. (2023). A survey on hallucination in large language models. arXiv:2311.05232. https://arxiv.org/abs/2311.05232
Xu, Z., Jain, S., & Kankanhalli, M. S. (2024). Hallucination is inevitable: An innate limitation of large language models. arXiv:2401.11817. https://arxiv.org/abs/2401.11817

Sam Altman. (2024). Reflections. https://blog.samaltman.com
OpenAI. (2023). GPT-4 technical report. https://arxiv.org/abs/2303.08774

法規與政策
European Parliament and Council. (2024). Regulation (EU) 2024/… on Artificial Intelligence (AI Act). Official Journal of the European Union.
White House Office of Science and Technology Policy. (2022). Blueprint for an AI Bill of Rights: Making automated systems work for the American people. https://www.whitehouse.gov/ostp/ai-bill-of-rights/
uropean Parliament, & Council of the European Union. (2016). Regulation (EU) 2016/679 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data (General Data Protection Regulation). Official Journal of the European Union. https://eur-lex.europa.eu/eli/reg/2016/679/oj
夏慧馨、陳志宇、鄭爲珊（2024）。通傳產業運用 AI 之法制議題初探。NCC NEWS，18 (2)。取自：https://newsweb.ncc.gov.tw/202406/ch1.html
中央研究院資訊科學研究所（2022）。台灣人工智慧倫理指引（AI Ethics Guidelines in Taiwan）。取自：https://ai.iias.sinica.edu.tw/ai-ethics-guidelines-in-taiwan

供應鏈管理、採購與談判理論
Monczka, R. M., Handfield, R. B., Giunipero, L. C., & Patterson, J. L. (2020). Purchasing and supply chain management (7th ed.). Cengage Learning.
Fisher, R., Ury, W., & Patton, B. (2011). Getting to yes: Negotiating agreement without giving in (3rd ed.). Penguin.
Nash, J. (1950). Equilibrium points in n-person games. Proceedings of the National Academy of Sciences, 36(1), 48–49.
Akerlof, G. A. (1970). The market for ‘lemons’: Quality uncertainty and the market mechanism. Quarterly Journal of Economics, 84(3), 488–500
Kraljic, P. (1983). Purchasing must become supply management. Harvard Business Review, 61(5), 109–117.
Vickrey, W. (1961). Counterspeculation, auctions, and competitive sealed tenders. The Journal of Finance, 16(1), 8–37.
Laffont, J.-J., & Tirole, J. (1993). A theory of incentives in procurement and regulation. MIT Press.
Bajari, P., & Tadelis, S. (2001). Incentives versus transaction costs: A theory of procurement contracts. RAND Journal of Economics, 32(3), 387–407.

賽局、行為心理學、LLM無法模擬人類心理
Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47(2), 263–291
Myerson, R. B. (1979). Incentive compatibility and the bargaining problem. Econometrica.
Harsanyi, J. (1967–1968). Games with incomplete information played by “Bayesian” players. Parts I–III. Management Science.
Moore, D. A., & Healy, P. J. (2008). The trouble with overconfidence. Psychological Review, 115(2), 502–517.
Loomes, G., & Sugden, R. (1982). Regret theory: An alternative theory of rational choice under uncertainty. The Economic Journal, 92(368), 805–824.
Binz, M., & Schulz, E. (2023). Do large language models learn human-like strategic preferences? Proceedings of the 45th Annual Meeting of the Cognitive Science Society.
Fudenberg, D., & Maskin, E. (1986). The Folk Theorem in repeated games with discounting or with incomplete information. Econometrica.