· Deployment & production· 日本語で読む
AgiBot’s 24-hour convenience store test with six G2 robots completes 1,730 tasks
A 24-hour convenience store livestream challenge involving six AgiBot G2 robots has ended. According to leaderobot.com, the six robots completed 1,730 tasks, including customer service and item pickup and delivery. The report gave the task count but did not describe human intervention or the conditions used to calculate success rates.
What happened
leaderobot.com reported on Nov. 3, 2026, that AgiBot’s 24-hour convenience store livestream challenge, billed as a test of whether “robots can keep working,” had ended. Six G2 robots completed a total of 1,730 tasks: 683 customer service interactions, 369 item pickup and delivery tasks, 147 coffee-making tasks, 126 breakfast service tasks, 277 restocking and patrol inspection tasks, and 128 tabletop cleaning tasks. The article also said four robots succeeded in “zhuanzheng,” but did not explain what the term means. The livestream drew more than 2 million viewers across the internet, it said.
Why it matters
The reported breakdown shows that the test covered several types of convenience store work, including customer service, item pickup and delivery, and restocking and inspections. Retailers assessing where automation can be applied need operating data that includes completion criteria for each task and human intervention, not just the total number of tasks. The article did not provide that information.
Analysis
Of the 1,730 tasks reported, customer service accounted for about 39.5%, item pickup and delivery for about 21.3%, and restocking and patrol inspections for about 16.0%. Coffee-making, breakfast service and tabletop cleaning were also included. The figures suggest the test involved different tasks across a store rather than a single-task demonstration. However, the reported task counts do not show how long each task took, how difficult it was or how many tasks each robot handled.
The notable aspect of the announcement is that it presented a 24-hour test period alongside categories ranging from customer service to cleaning. But the total of 1,730 completed tasks alone does not reveal the success rate against all attempts or how much human intervention occurred. Assessing the results as evidence that robots can autonomously sustain store operations would require additional information, including task definitions, how failures and retries were counted, and staffing during operation. The article also did not explain “zhuanzheng,” so it is not possible to determine from this report what qualified the four robots for that status.
A related article from this newspaper, “Gizmodo Japan reports on 24-hour convenience store test using AgiBot’s G2”, identifies test conditions, human intervention and how results were measured as points to verify. The task counts in this report, like those in the related article, provide information about testing in a store environment, but should be read as a different metric from operating hours or model evaluations. In addition, [“AgiBot showcases operating cases at import expo; report says cumulative effective intelligent working hours exceed 62,000”](/en/news/agibot-showcases-operational-cases-at-import-expo-reported-effective-i-bfc739) reported a cumulative figure as of October 2026. Cumulative operating hours and the task count in this test differ in both scope and meaning and cannot be substituted for one another.
For those deciding whether to deploy robots in retail, a key question is how tasks such as customer service, item pickup and delivery, and restocking and inspections can be combined within a store’s operations. However, the information confirmed by this article is limited to task counts and livestream viewership; it does not substantiate repeated operation in a real store, staffing reductions or profitability. The next points to verify are task-by-task completion criteria, records of intervention and failures, the conditions of the 24-hour operation, and the definition of “zhuanzheng” for the four robots.
Sources
Translated from the Japanese original: /news/agibot-g2-robotto-6-dai-no-24-jikan-konbini-shiken-wo-houdou-e9fc14. The Japanese version is authoritative if they differ.