The 12th Last Mile Delivery Conference was held in Beijing on November 6, 2026, under the theme "Transforming Toward Innovation, Warming the Last Mile." Zhang Hao, Innovation Business Lead of JD Express, attended and delivered a speech. The following is the transcript of the speech:

Hello everyone, I am Zhang Hao from JD Express. Below I will report to all the leaders on some innovations and developments in smart unmanned equipment.

Warming the Last Mile: The Core Premise of Smart Equipment

Today's topic is "Warming the Last Mile." This "warming" has two layers of meaning: first, warmer service; second, warmer feelings for the delivery riders. Back to daily work, the topics we discussed this year are nothing more than village entry, door-to-door delivery, cost, and revenue — every company is probably similar. But these topics have a certain contradiction with "warming the last mile": all our work must ultimately fall on the last-mile riders to complete, and after this work is converted into workload, it actually goes against reducing the overall labor intensity of the riders. Against this backdrop of the times, what exactly are we doing? It is nothing more than the balance of cost, efficiency, and experience. In this process, we have already done some things, such as the application of AI technology, more optimized allocation algorithms, and more human-friendly guided operations and other software improvements. But when it comes to the riders moving goods, labor is still being expended. We want the riders' daily parcel count to keep rising and bring more income — through management methods we can push them from 100 parcels to 200, and someday from 200 to 300, but where is the ceiling of this? How to better increase capacity is actually a development path more in line with human nature and future trends. This actually comes down to the transformation of operating models, and how the operating model transforms is closely related to smart equipment. Today when we talk about smart equipment, we are not simply talking about the growth of the equipment itself, but about how it brings model transformation within the express logistics industry, and how we reverse-engineer the continuous improvement of smart equipment under model transformation. The overall big development of the robotics and physical AI industry this year has given us more opportunities and chances. The data and views just mentioned, I believe everyone present including myself already had some consensus and shared understanding on before, and today's sharing by the previous guests also reinforced this. If we want to make it easier for riders to move boxes, we indeed have some better opportunities. Developing unmanned and smart equipment has one big premise: we are not trying to replace the riders, but to hand over pure physical exertion and labor to robots, and leave the "warmth" to people. This is the core major premise and principle for our development of innovative businesses and smart equipment.

This Year's Development Opportunities: Technology, Scenarios, and Policy

(1) Technological breakthroughs. This year's technological development has seen breakthroughs at several levels across different smart equipment industries. For example, in the unmanned vehicle industry, safety has improved this year and BOM costs have declined — as the previous guests also shared, the commercial cost chain of unmanned vehicles has already been proven, and cost efficiency has been validated. For example, for drones, flight control capabilities have also improved through this year's investment and development, but the road ahead may be longer than that of unmanned vehicles. For example, for building robots, elevator control technology has been solved this year under the innovation trend, so building robots truly have applications in communities and cities, not just in office buildings and high-end residential compounds. Different categories of smart equipment have indeed made considerable progress.

(2) Scenario expansion. Scenarios keep advancing, including village entry and the upward flow of agricultural products, which have brought great opportunities for smart equipment. For example, the leader from Guizhou earlier shared how to enter villages and get agricultural products out. JD has done the same — we have also found some scenarios and locations where the capabilities of smart equipment have been fully demonstrated.

(3) Policy opening. This year, under the improvement of safety and the guidance of the overall "baton," policies have opened up to a certain degree, giving companies like JD some development opportunities and chances.

JD's Practice: Full-Chain Intelligence

Looking at the intelligence of the entire chain, we have never taken express delivery out separately as the main battlefield, but instead viewed the entire logistics supply chain as a whole and pursued overall intelligence. In earlier years, JD's accumulation inside the warehouse "within four walls" laid a solid foundation for transportation and last-mile scenarios, so we quickly built a complete set of scheduling capabilities and coordination capabilities for different types of robots. Starting from relatively mature in-warehouse automation, we proactively extended outward to the transportation layer and the touchpoint layer, focusing on building scenario-based capabilities at these layers, and ultimately hoping to create an innovative business map covering the entire chain.

(1) Ground: Cargo unmanned vehicles. Cargo unmanned vehicles have made considerable progress. Not only are these products built by JD itself, we have also integrated and scheduled socialized express unmanned vehicle resources and capabilities. Around unmanned vehicles, our capability scenarios have already begun to cover: in addition to unmanned vehicle coverage in express delivery scenarios, we have also made attempts in other scenarios, including direct dispatch from sorting, store business, and innovative retail businesses within the JD ecosystem, with businesses such as Jingxitong all making attempts in different scenarios. In addition, we have also developed same-city patrol capabilities: on the one hand providing idle-time capacity for society to share, and on the other hand linking with firefighting, transportation, and public security departments, making patrol unmanned vehicles a tool and eyes for social management.

(2) Air: Drone network. In terms of aerial unmanned operations, JD has also made some different attempts. Among the more innovative ones, for example, direct dispatch from the warehouse end: after JD completes outbound shipment, the drone flies directly to the consumer's hands. There is also the rural scenario: in the southwest, we have built the largest drone express network covering different township areas.

(3) Air-ground coordination: Practices in Suqian and the Southwest. Taking Suqian as an example, our unmanned vehicles and drones have carried out some strong coordination: drones can land on unmanned vehicles, and we also have drone operating capabilities at unmanned vehicle patrol bases. Many frontline riders come from the express logistics industry or the food delivery industry. They are both warehouse managers or couriers, and also drone pilots and unmanned vehicle safety officers — this actually provides them with different employment directions and is also a plan for fully utilizing human resources. At present, we have about 25 normalized drone routes in Suqian, covering Suqian's airspace, and are carrying out new scenarios of unmanned vehicle and drone coordination. In the southwest region, we have more suitable scenarios and more open airspace policies, covering rural scenarios mainly in some county towns, and achieving daily village entry for e-commerce and logistics. The routes we selected are all scenarios with major rivers or mountains as natural barriers, where an overall drive of more than one and a half hours is shortened to a flight of about 8 minutes. This brings more than just an experience improvement. There is one data point: in some county areas covered by drones, after the coverage network was built, local e-commerce order volume increased by more than 30%, which is a typical case of e-commerce benefiting from experience. The upward flow of agricultural products has also been well validated through drone links. Overall, the returns made by drones in specific scenarios are completely feasible, and there is still great room for cost reduction and technological development — perhaps in the next 3 years or after some time, its development heat will be like that of current unmanned vehicles, and this judgment is basically consistent.

Next Steps

In complex scenarios, we have already begun to do full-chain linkage, opening up the automation capabilities of JD's overall supply chain, achieving full-chain automation in some regions, so that riders across the entire e-commerce logistics supply chain industry can feel the cost reduction, efficiency improvement, and labor intensity reduction brought by physical AI, and these will truly be reflected in their income. To sum up, JD's innovation in smart equipment has not remained at the theoretical level, but has focused on implementation and rationalized cost input-output. We have not blindly pursued a certain scale, but have made reasonable attempts on the premise of improving warmth and riders' income, making the entire attempt stable and sustainable. Through smart equipment, we have achieved several effects: first, greater stability and predictability in experience — of course, this also depends on more stable airspace policies and right-of-way policies; second, cost reduction, efficiency improvement, and enhanced resilience of the entire network; third, social mission — in addition to the social security dimension and the dimension of riders' rights and interests, we also hope that through such methods, some blue-collar workers can achieve true transformation, and in the future provide them with better jobs and more technically sophisticated career choices. Next, we will do several things. First, in the continuous construction of rural capabilities, rely on smart equipment to make further breakthroughs and expand the network. Second, make smart equipment more reliable. Here is one piece of information worth sharing: when express unmanned vehicles were first put into the industry, the monthly availability rate might be less than 90% — considering right-of-way restrictions, weather factors, road construction, and other actual conditions. To make the business model and operating model more stable, it is necessary to make the equipment safer and more reliable, for example, making express unmanned vehicles as stable as PDA operating software, reaching a 99.9% availability rate, so that we can truly implement the operating model and truly bring about model transformation and innovation. Third, we will also proactively open up our overall capabilities to the ecosystem, so that physical AI technology can benefit more practitioners. That is all for my sharing, thank you.