The latest experiment of AI plus delivery workers, 2026 Meituan and Ele.me use AI to reshape the delivery industry in 5 major directions

Starting from the second half of 2025, Meituan, Ele.me, and JD Daojia will successively embed AI into every aspect of takeout delivery. The rider app has a built-in AI assistant that plans routes in real time, the merchant's AI automatically takes orders and distributes meals, and the background AI dispatches millions of orders. This article tells you from the 5 directions of actual implementation in 2026, what exactly AI plus delivery drivers are changing, whether the income of ordinary riders has increased, and whether the delivery experience of ordinary users has improved. This article does not cite specific percentages and absolute numbers that fluctuate greatly, but only discusses directional changes.

Direction 1, intelligent path planning

Riders used to rely on experience and maps to plan food delivery routes. During a peak period, the number of orders may be overwhelming. Which one to send first depends entirely on memory and intuition. Detours are common.

Meituan and Ele.me have successively launched intelligent routing assistants based on real-time traffic, order timeliness, merchant delivery speed, and user location. The rider app directly displays the recommended route, which can save a considerable amount of time on average for each order. The specific improvement rate varies greatly depending on urban road conditions, and is subject to the official public data of the platform.

Even smarter is dynamic adjustment. If a new order is received on the way or a certain merchant is slow to deliver food, the system will automatically recalculate the route and respond almost in real time.

Actual benefits: The average daily delivery volume and working hours of riders in the pilot cities are moving in the direction of "increasing order volume and staying the same or declining". This is the most direct benefit of AI path planning.

Direction 2: Merchant-side AI takes orders and distributes meals

Merchant-side AI tools will be launched one after another in the second half of 2025. It does three things.

The first is order consolidation. If there are multiple orders from the same community, the system automatically prompts the rider to pick up the food at once, reducing the rush.

The second is the meal estimate. Based on order complexity, kitchen load, and historical data, it predicts the delivery time of each order and tells the rider "I'll pick up the meal in a few minutes" or "Wait a little longer."

The third is abnormal warning. If a merchant's multiple orders in a row time out, the system will automatically prompt the rider to postpone taking new orders from the merchant, or notify the dispatch center.

Pilot merchants generally reported an increase in on-time meal delivery and a decrease in complaints from riders waiting for meals. The specific figures will be subject to a joint review and disclosure by the platform and merchants. This is a win-win situation for merchants, riders and users.

Direction three, intelligent dispatch center

The dispatch center is the central nervous system of the platform's backend. In the past, it relied on manual work and simple algorithms. Starting from 2025, mainstream platforms will be upgraded to AI-driven scheduling systems.

Scheduling AI does three things: order matching, which finds the optimal rider for new orders within a few seconds based on the rider's location, load, and ability; supply and demand adjustment, which predicts a surge in orders in a certain area and automatically sends dispatching fee rewards to attract riders; exception handling, which automatically redistributes orders without affecting the overall timeliness of abnormal situations such as rider illness, vehicle failure, and heavy rain road closures.

The scheduling time-consuming and order timeout rate indicators have been significantly improved. The specific extent of improvement is different from the public data of each platform, and the official disclosure shall prevail.

Direction 4, rider AI assistant

Rider personal AI assistant is the most imaginative direction in 2026. In addition to path planning, it can do three things.

The first is revenue optimization advice. AI analyzes the rider's historical order data and recommends which time period and area to receive the highest profit. Top riders report that there is room for improvement in monthly income.

The second is health monitoring. The Rider App can access heart rate watch data, the system monitors the rider's status, and automatically recommends a break after continuous high-intensity delivery for a period of time.

The third is training assistants. After new riders join the company, they will be guided one-on-one by an AI assistant to answer questions about delivery rules, merchant distribution, and customer preferences. The level of familiarity will increase significantly faster than in the past when the old rider was simply brought in to guide the new rider.

The feedback from the rider community is generally positive, and the majority of riders are willing to continue using AI assistants. The specific satisfaction ratio is based on community surveys.

Direction five, AI plus unmanned delivery

This is the furthest direction but is already being piloted. Meituan’s self-driving cars have been piloted in Beijing, Shenzhen, Guangzhou and other places in the past few years. The specific number and proportion of orders covered are subject to official disclosures.

The AI ​​system of unmanned vehicles consists of three modules: autonomous navigation, cargo identification, and exception handling. However, unmanned vehicles will not replace riders on a large scale in the short term. There are three limitations: the cost of the entire vehicle is still high, and depreciation costs are not necessarily superior to labor. Due to road conditions, they can only run in closed scenes such as parks, campuses, and university towns, and there are still challenges on urban roads. Delivery is upstairs, and unmanned vehicles are sent downstairs. The last 50 meters still need to be carried by the rider or a locker.

It is expected that for many years to come, autonomous vehicles will collaborate with riders: autonomous vehicles run long-distance main roads, and riders run the last mile.

Will AI increase or decrease rider income?

Riders at different echelons have different feelings.

First, the income of top riders has increased. With AI path planning and AI income optimization suggestions, the monthly income of leading riders has a relatively stable room for improvement. The specific extent varies greatly depending on the city and order density.

Second, the median rider’s income remains the same. The average daily order volume increased, but the platform unit price dropped slightly, and the comprehensive income fluctuated little.

Third, new riders have improved significantly. The AI ​​training assistant allows newcomers to get started quickly, and the first-month income is significantly higher than the same period in the past when relying on mentorship.

Overall, AI has increased the income differentiation among rider groups. Riders who adapt quickly will earn more, while riders who cannot keep up with the pace will still have a hard time.

Career changes of delivery workers in the AI ​​era

Three obvious changes.

First, working hours tend to shorten. In the past, the average daily working hours of riders were generally longer. After the introduction of AI, the output per unit time has increased, and there is room for compression of the total time while maintaining the order volume.

Second, the accident rate has dropped. AI route optimization and health monitoring allow riders to reduce fatigue driving. The delivery accident rate in major cities has declined in recent years. The specific data is subject to the public statistics of the traffic management department.

Third, there are more transformation options. Riders can choose to become AI dispatchers, trainers, operations specialists, and merchant BDs. The demand for these positions continues to increase. Riders are no longer a "no-growth" profession, AI has opened an upward path for riders.

What changes do ordinary users feel?

Three changes.

First, delivery time is shortened. The average delivery time in major cities has dropped significantly in the past two years. The specific minutes are based on the platform's annual public report.

Second, punctuality is improved. The on-time performance data of Meituan and Ele.me continue to rise.

Third, exception handling is fast. When there are problems with orders, such as merchants missing meals or riders delivering to the wrong address, the AI ​​customer service response time is significantly shortened, and the average processing time is significantly shorter than in the past. User satisfaction with the takeout experience is improving overall.

The next step in AI-based food delivery in 2026

Three directions to be broken through.

The first is the pre-order AI. Based on the user's historical orders and current life rhythm, it prompts the user in advance that "it's time to order lunch", which is already being tested.

The second is health AI. Analyzing dietary structure based on user orders and proactively recommending healthy options involves privacy disputes.

The third is merchant product selection AI. Based on user preferences in this region, we help merchants automatically adjust their menus, which is currently being piloted on a small scale.

Overall, AI combined with food delivery is one of the directions with the greatest potential for scale in the local scene of AI in China, affecting riders, merchants, and users, and is a structural upgrade rather than a simple replacement for the job market.

FAQ

Will the AI ​​food delivery system put riders out of work?

There will not be large-scale unemployment in the short term, but some will be replaced in the long term. The core value of a rider is the last mile plus emotional labor, these two points cannot be replaced by AI in the short term. However, simple order merging, order inquiry, and customer service links will be replaced by AI, which account for a large part of the rider's working time. After autonomous vehicle technology matures, closed scenes such as university towns and office parks will replace some riders, but riders will still be needed on urban main roads. It is comprehensively predicted that the total number of riders will shrink to a certain extent in the medium to long term, but it will not collapse significantly.

Is AI scheduling fair? Does it favor certain riders?

Platform officials say the algorithm is fair to all riders, but the rider community reports that there are hidden biases. Three types of riders have higher actual benefits: riders with high acceptance rate, riders with high punctuality rate, and riders in new areas. This bias is efficient and reasonable but unfair to veteran riders, causing controversy in the community. Some city regulators have required platforms to disclose algorithm logic and promote algorithm transparency.

What data does the food delivery AI assistant collect?

The collection is huge. Location data, order history, delivery tracks, heart rate watch data (if connected), all click records in the rider app, and conversation records with customers will be collected. In addition to optimizing the AI ​​algorithm, these data are also used for rider profiling and grading. Starting from 2026, the "Personal Information Protection Law" will have stricter requirements for platforms to collect rider data, and platforms must clearly indicate the purpose and retention period of data.

Will AI make takeout cheaper?

It has not become cheaper in the short term, but it may decline in the long term. In the short term, the average takeaway price will still rise along with overall commodity prices. However, the improvement in AI scheduling efficiency has reduced platform operating costs. This part of the money saved is currently mainly given to the platform and shareholders, and has not been fully transmitted to consumers. In the long term, after unmanned delivery is rolled out on a large scale, delivery fees are expected to drop further, which may lower the final price.

Does AI make it harder or easier for riders?

The data is easier, but the actual experience is more complicated. Looking at the data, working hours have been shortened, accident rates have dropped, and income brackets have improved, all of which point to better outcomes. But the actual experience of riders is more complicated: AI monitoring makes the "fishing" space almost zero; the AI ​​algorithm continues to optimize, and yesterday's best path will be outdated today, so you have to stop learning; AI makes orders more intensive, and although the total duration is reduced, the intensity per unit time is higher. Overall, riders who adapt to the AI ​​rhythm feel more relaxed, while riders who cannot keep up with the rhythm feel more pressure.

Source of inspiration: Issue 386 of Ruan Yifeng's "Technology Enthusiasts Weekly" https://www.ruanyifeng.com/blog/2026/02/weekly-issue-386.html