[1M5-GS-10] AI application:

[1M5-GS-10-05] Development of the AI Agent for the Financial Negotiation With Deep Reinforcement Learning

〇Atsumi Saito¹, Ryota Higa¹, Satoshi Morinaga¹ (1. NEC)

Keywords:AI Negotiation, Finance, Deep Reinforcement Learning

Financial products always have multiple condition items, requiring substantial time for condition adjustments with customers. This study focuses on developing an AI agent to automate negotiations with customers. By using a simulator that replicates customer behavior as the environment, we train the agent through deep reinforcement learning to propose conditions as a product provider with customer's requests. The results showed that the trained agent significantly improved agreement rates with customers and proposed more profitable conditions per contract compared to the initial agent making random proposals. This suggests that deep reinforcement learning can generate AI agents capable of achieving better negotiations with humans.