徐启昌：70%-80%的大模型项目投资回报未达预期

Core Insights - The industry is undergoing a comprehensive transformation, shifting from a "seller product sales" model to a "buyer advisory" model, focusing on customer lifetime value and comprehensive solutions [2][7] - Customer coverage is expanding from high-net-worth individuals to a broader audience, including middle-aged, younger generations, senior citizens, and rural populations [2][7] - Customer demands are becoming increasingly diverse, encompassing both stable investment needs and high-risk, high-reward aspirations, as well as extending to family inheritance scenarios [3][7] - Product innovation is accelerating, with banks not only enriching their own product systems but also introducing public funds and insurance as part of their ecosystem [3][7] - The logic of technological support is evolving, with AI becoming a core production tool that reconstructs the entire business process [4][7] Industry Challenges - There is a contradiction between conservative regulatory policies and limited application scenarios for technology, as AI is currently not allowed to directly replace human management in trading, which restricts the full release of technological value [4][7] - Approximately 70%-80% of large model project investments have not met expectations, primarily due to discrepancies in model selection, application scenarios, and implementation methods [4][7] - The recommendation for institutions is to adopt a "small steps, quick iterations" approach to enhance investment returns [4][7] Data Security and Compliance - There is an optimistic view regarding balancing data security and compliance, suggesting that technological means can effectively resolve the contradictions between regulatory compliance and innovative development [8] - Techniques such as pre-processing constraints and post-processing checks can prevent data leakage, while privacy computing and data de-identification can achieve data usability without visibility [8]