Many MindedShanghai Convenience Store Trial Shows AGIBOT Robots Handling 1,730 Retail Tasks in 24 Hours
Six AGIBOT G2 robots performed 1,730 tasks during a 24-hour trial at a convenience store in Shanghai’s Xuhui District, running from 10:00 November 2 to 10:00 November 3, 2026. Four robots passed a performance assessment, while two did not meet all criteria. The robots completed 683 customer-service interactions, 369 product pickups and deliveries, 147 coffee orders, 126 breakfast servings, 277 shelf inspections, and 128 table-cleaning tasks during the trial. The trial tested embodied AI capabilities in real-world retail operations using AGIBOT’s Genie Master framework, which coordinated robot actions and learning from operational feedback. Robots demonstrated ability to distinguish similar products (e.g., regular vs. sugar-free Coca-Cola), rearrange items to correct shelf positions, and navigate around obstacles like people and boxes.
This trial assessed robots on time on duty, task completion, customer feedback, response times, and handling unexpected situations. The metrics reflect performance during this specific 24-hour period across four operational cycles: lunch rush, evening peak, overnight, and morning breakfast. While the robots showed functional capabilities in retail tasks, the trial represents a single operational test—not scaled production deployment.
This demonstrates embodied AI’s potential to reduce labor-intensive retail operations at scale. If the Genie Master framework achieves consistent performance in real-world conditions, it could lower costs for convenience store labor. However, current results are limited to a single trial with unverified thresholds for passing assessments. Next steps include scaling trials and confirming whether the robots maintain performance beyond this test period. The source explicitly states the trial was not a production deployment, and exact assessment criteria beyond pass/fail status remain unspecified.
Source: The Robot Report