Evaluation of Vertex AI Agent for Investment Product Support: Enhancing Customer Service in Asset Management

This study explores the application of a Vertex AI agent designed to support asset management customers by providing detailed information about investment products through a natural language interface. We discuss the agent's architecture, which integrates an advanced language model and API, enabling user-friendly access to data on asset classes, performance metrics, fees, risks, and liquidity. The paper evaluates the agent's accuracy, error-handling capabilities, and iterative improvements, illustrating how targeted training enhances the agent?s performance in responding to customer inquiries. Future improvements for addressing calculation-based queries and data fabrication issues are also proposed.