In order to actively carry out the three-year action plan of China Post Group for AI applications, the Parcel, Express and Logistics Business Unit has been committed to establishing its own large language model (LLM) for the specific scenarios of parcel, express and logistics business, and has completed the private deployment of the DeepSeek-R1 recently. This marks that the business unit has officially embarked on a new journey to the intelligent applications of generative AI to embrace the “All In AI” era.

Since it initiated the “artificial intelligence plus” transformation in 2020, the business unit has elaborately designed the interaction process between intelligent robots and customers through in-depth applications of AI engine, and created a line of products including intelligent voice navigation, intelligent text-based customer service, and intelligent assistants. It has also upgraded its 11183 Customer Service Center into an intelligent interaction center, thereby constantly empowering customer service to improve quality and efficiency.

To avoid the waste of assets resulting from hasty investment in computing power resources in the context of the rapid development of LLM technology, the business unit actively explored LLM applications tailored to its specific needs, and carried out LLM data training and performance comparison under its own peculiar scenarios in 2024, based on the intelligent architecture of domestic LLMs. During the Spring Festival in 2025, the emergence of DeepSeek, with its superior performance and low resource dependence, provided an opportunity for the rapid construction of the scenario-specific LLM of parcel, express and logistics business. The business unit responded promptly, and connected to DeepSeek-R1 as soon as possible. It then conducted pre-research on scenario-specific applications, pushed forward functional testing, performance optimization and security validation, and expanded pre-training and fine-tuning efforts of the scenario-specific LLM, thereby laying a solid foundation for practical scenario applications.

Next, the business unit will deeply integrate DeepSeek with its business scenarios, and give full play to its capabilities of cognitive reasoning, intelligent innovation and scenario-specific applications to reshape work process. It will develop various scenario-specific assistants such as intelligent customer assistants, intelligent customer service agents and digital employees, thus enabling round-the-clock business transactions by AI robots. With these efforts, the overall customer service efficiency in some scenarios is expected to increase by more than 33 percent, bringing customers more convenient, efficient and intelligent service experience.