Aishida’s announced 100-store humanoid rollout entered a three-day public retail test on September 28, with customer-facing robots scheduled to work from 10 a.m. to 8 p.m. through September 30 in Wenling, Shanghai, and Wuxi. Cailianpress reported on the trial while it was underway, including a September 28 operating scorecard. The format matters: the available evidence documents a public livestream and test window, and it does not establish sustained operation at every site or a completed measure of the program’s retail value.
The retailer is cookware company Aishida, identified as ASD in AGIBOT’s English announcement, and the supplier is robot maker AGIBOT, known as Zhiyuan in Chinese-language coverage. Shobserver reported that 100 Lingxi X2 units began entering Aishida’s 100 stores on September 28, while AGIBOT calls the same unit AGIBOT X2. The three-day test keeps the model names and program scope tied to the terminology used by the participating companies and reporting.
A 100-store humanoid retail program runs across three cities
The announced deployment places 100 humanoid robots across 100 Aishida retail stores in China. That one-to-one framing is part of the announced rollout, but it does not establish permanent use or the same level of activity at every location. AGIBOT’s September 28 announcement and Shobserver’s reporting establish the reported count, store footprint, dates, cities, and public schedule. They do not provide an independently audited account of fleet utilization after the livestream period.
The public program concentrated its activity in Wenling, Shanghai, and Wuxi, where the broadcast ran for 10 hours per day from September 28 through September 30. The locations supplied a visible setting for a retail exercise that put the robots in front of shoppers during the announced public broadcast. Cailianpress, Shobserver, and AGIBOT describe the dates, cities, and customer-facing roles, but not a longer operational period beyond the three-day event.
The deployment’s naming requires care. AGIBOT uses AGIBOT X2 in its English-language announcement, while Shobserver uses Lingxi X2, the Chinese name connected to the same Aishida work. Treating those labels as two different models would overstate the evidence. Conversely, adding a suffix or substituting another AGIBOT product would create a false precision absent from the accounts of this test.
Aishida keeps commercial decisions with human associates
The reported job design split customer interaction from the commercial steps that complete a purchase. The robots were assigned to greet customers, ask about needs, introduce products, recommend cookware, and guide shoppers in stores. Promotions, inventory, and the final purchase remained with Aishida’s human sales associates, according to AGIBOT. Shobserver likewise described a handoff to a human guide for offer details and completion, placing the retail roles within a human–robot workflow that did not encompass an end-to-end automated sale.
One livestream interaction illustrates that division without establishing a fleet-wide performance result. Shobserver’s reporter described a robot asking a shopper about household size, cooking preferences, and budget, then producing a cookware recommendation in about 30 seconds. Promotion details then moved to a human associate. AGIBOT presented a similar customer example in its own announcement. The account supports a bounded picture of needs assessment and recommendation, not a claim that every robot, store, or customer followed the same sequence.
EastFrontier has also covered Aishida’s planned retail-store trial and AGIBOT’s theme-park deployment. They are separate archive reports and do not substantiate the results of this three-day retail test.
AGIBOT’s published description of the program also treats store employees as part of the operating arrangement. Its account assigns staff the promotion, inventory, and purchase steps after the robot’s initial interaction. Zhihui Peng, AGIBOT’s Co-founder, President, and CTO, characterized retail settings as requiring robots to work with varied customers, respond to different needs, and operate alongside store employees throughout the day. That company framing aligns with the reported handoff, but it does not establish a retail outcome beyond the test itself.
For Aishida, the robot’s reported role was customer-facing assistance, while promotions, stock, and purchases remained with people in the store. The reported deployment therefore describes a handoff between the robot’s interaction tasks and associates’ commercial tasks.
Interaction counts leave retail economics unanswered
Cailianpress published a September 28 scorecard for three livestream stores in Wenling, Jiangsu, and Shanghai: 10 hours of operation, 74,080 steps, and 4,299 human-robot interactions. Those figures indicate the kinds of operating measures the public test placed in view, but they must remain attributed to Cailianpress and cannot be treated as audited measurements. Cailianpress does not describe a measurement method, and the published totals do not independently establish uniform activity across the full 100-store announced rollout.
Interaction volume is not a disclosed retail result. Douchuang/Sohu reported that the launch materials did not disclose sales-conversion rates, changes in labor input, or deployment costs. The available record documents what the units were assigned to do and what was publicly counted, but it does not provide a retail-economics result for the three-day test.