Automation to intelligence

Over the past year, major courier companies, including SF Express, ZTO Express and J&T Express, have integrated large language and industry-specific AI models into their operations, marking a transition from basic automation to more adaptive, system-wide intelligence.

At J&T, AI-driven training and internal support systems have reduced manual customer-service workloads by about 50 percent and shortened onboarding time for new employees. At SF and ZTO, large models are being used to optimize resource allocation, predict demand surges and streamline service workflows, helping lower operating costs while improving reliability.

AI has been embedded in nearly every major link of the delivery chain. Vision systems scan parcels in milliseconds, reducing sorting errors. Algorithms adjust routes in real time in response to weather, traffic and volume fluctuations. Predictive models help companies anticipate peak demand during shopping festivals and holiday travel seasons.

By the end of China's 14th Five-Year Plan (2021-25), most large sorting centers had achieved near-total automation. Cloud computing and large models were being deployed for demand forecasting, warehouse management and dispatch planning, while electronic waybills and reusable transit packaging became standard across the industry, Zhao said.

Inside the warehouse

At JD Logistics' Pingshan grid warehouse in Shenzhen, the transformation is visible on the floor.

The site processes roughly 15,000 parcels a day, with five unmanned vehicles shuttling goods on site, reducing repetitive trips by human staff.

"They ease the workload and let couriers spend more time in communities, focusing on customer service," said An Jixing, the station's head. "The improvement in efficiency is very clear."

Nearby, JD's "smart warehouse" that opened in October, uses nearly 200 mobile robots and vertical transport systems to move standardized bins through a dense, three-dimensional storage environment. It can handle more than 35,000 outbound orders a day.

According to Li Jinyuan, an operation and maintenance engineer at JD Logistics' smart warehousing facilities, parcels can move from order confirmation to outbound sorting in as little as 15 minutes. Storage efficiency is more than four times that of traditional warehouses, while picking accuracy has reached 99.99 percent, he said.

As machines have taken over repetitive tasks, new roles have emerged to manage, maintain and optimize intelligent systems.