Midea’s washing machine factory in Jingzhou has received certification from the World Record Certification Agency (WRCA) as the “world’s first artificial intelligence agent factory with multi-scenario coverage.”
As the first of its kind in the industry, the plant runs on a system it calls a “factory brain,” which orchestrates 14 AI agents across 38 business scenarios. These span functions from perception and decision-making to execution, feedback, and continuous optimization. In practice, the agents are said to complete tasks that once took hours of manual work in seconds, raising efficiency by more than 80% on average. Scheduling response times are up 90%.
The system acts as the plant’s “nervous system.” Built on a distributed, scalable multi-agent architecture, it enables AI agents to work autonomously through agent-to-agent (A2A) communication and enhances decision-making with industrial large model inference engines.
If the factory is a body, the brain is its task dispatcher, the AI agents are its neural networks, and intelligent terminals such as robotic arms and sensors are its limbs.
One common task at the Jingzhou facility is quality inspection. Workers wear an AI-powered headset linked to the system, which highlights common error points based on market data and historical inspection results. Using vision technology, the headset connects with both R&D and quality control agents to automatically retrieve blueprints, compare them with captured images of materials, and return inspection results. This has reportedly cut initial check times from 15 minutes to about 30 seconds.
At an automatic fastening station for dryer rear covers, a Kuka robot works with a planning AI agent. Real-time image uploads allow the system to identify product features and match them with multiple washing machine models. Even on mixed production lines, the system can instantly adjust screw-tightening programs, enabling flexible, precise operations comparable to human adaptability.
Zhang Xiaoyi, Midea’s chief digital officer, said that under the system’s coordination, every production element—humans, machines, materials, methods, and environment—works as part of an interconnected whole. Intelligent terminals such as humanoid robots, robotic arms, AMRs (autonomous mobile robots), injection molding machines, cameras, and sensors are all equipped with perception, comprehension, decision-making, and action capabilities.
The humanoid robotics sector is growing quickly and is expected to play a central role in factories of the future. Unlike many manufacturers still in the conceptual stage, Midea has deployed humanoid robots in real-world industrial settings. At the Jingzhou plant, one such robot has completed staged testing in an injection molding workshop and been connected with factory business systems.
Midea’s AI research institute has developed Miro, a humanoid robot that serves as a core execution unit of the AI agent system. Directed by the factory brain, Miro works with quality control, DMS (daily management system), TPM (total productive maintenance), and EHS (environment, health, and safety) agents. With multimodal sensing and embodied operation technologies, the robot autonomously carries out frequent tasks such as quality checks, DMS meetings, TPM inspections, and EHS patrols, making real-time decisions across workflows.
In one example, Miro conducts an initial inspection by moving an injection-molded washing machine drum to a smart inspection station. Its self-developed 3D camera measures dimensions and appearance. The quality control agent receives the results in real time, generating a closed-loop work order. Only once a part passes inspection does the robot return to collect it. If defects are found, the system coordinates with process and TPM agents to fine-tune injection machine parameters.
Still, humanoid robots remain in the exploratory phase of large-scale application. Xu Yi, head of Midea’s AI research institute, said the top priority is scenario-based problem-solving. A greater variety of use cases provides the testing ground needed to advance technology and drive system evolution. Within the factory, multi-agent collaboration continues to strengthen the capabilities of robots and intelligent terminals, significantly boosting efficiency.
According to Zhang, Midea plans to expand the number of scenarios covered by AI agents and replicate the factory model across its global manufacturing network.
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This article was written by Huang Nan for 36Kr.