Taiwanese e-commerce platform momo Fubon Media announced a partnership with Microsoft Taiwan to develop a next-generation intelligent customer service system powered by a Large Language Model (LLM), which officially launched in July of this year. This solution, combining Microsoft Azure OpenAI and Retrieval-Augmented Generation (RAG) technology, significantly improves semantic understanding and instant response performance, achieving an accuracy rate exceeding 90%, marking a significant milestone for momo's journey towards becoming a "tech-enabled e-commerce company."
The Technological Leap from Chatbot to LLM
Since launching its smart customer service platform using chatbot technology in 2017, momo has continuously refined its offerings. Initially providing basic lifestyle and shopping information, it now supports order inquiries. Smart customer service has become a key service portal, with nearly 330 million users in 2024 alone. The upgrade to LLM AI customer service further demonstrates the culmination of years of development.
Fubon Media Technology General Manager Gu Yuanhong said, "This new generation of intelligent customer service not only brings a better customer experience and higher service efficiency, but also marks a new chapter for Momo in generative AI and deepening data applications. AI does not replace services, but rather enhances the understanding and immediacy behind them."
Improved technical architecture and service efficiency
momo has integrated Microsoft Azure OpenAI services, combining OpenAI's GPT model with the security and scalability of the Microsoft Azure platform to enable AI customer service to demonstrate enhanced semantic understanding and natural conversational performance. By incorporating a search-enhanced generation framework, the system ensures that responses are based on clear evidence, effectively overcoming the challenges of the LLM model's limited use of Traditional Chinese corpus and proprietary knowledge.
Since its launch in July, the new generation of LLM AI customer service has significantly improved its semantic understanding and response capabilities, with an accuracy rate of over 90%. Furthermore, customer willingness to use AI customer service self-service has increased by 5%, equivalent to a 5.7% increase in AI customer service staff, effectively reducing the workload of real customer service representatives.
Future development and application planning
Microsoft Taiwan General Manager Bian Zhixiang stated, "The e-commerce market is entering a new era of generative AI. We will work with momo to explore more AI application possibilities, extending from customer service to various aspects of e-commerce services."
Over the next two years, momo plans to further develop its intelligent customer service system, upgrading it to an AI shopping advisor as the core entry point for comprehensive customer service. It also plans to introduce emotion recognition capabilities, allowing the system to instantly perceive customer emotions and provide more engaging interactions. Furthermore, the MCP (Model Context Protocol) will facilitate cross-system integration and expand its retail application scenarios.
Through its technical collaboration with Microsoft, momo is accelerating AI-driven operational innovation, creating higher value for consumers, brand partners, and merchants, and leading Taiwan's e-commerce industry into a new era of AI.
