Xiaomi has completed a major internal restructuring of “Xiao Ai,” its flagship AI assistant that has been in operation for nearly a decade, according to reports citing multiple sources. The overhaul is designed to move Xiao Ai away from functioning as a standalone product team and toward becoming shared AI infrastructure across Xiaomi’s broader “Human x Car x Home” strategy.
Under the changes, Xiaomi split Xiao Ai’s technical architecture into three independent modules: a foundation model layer, a cloud layer, and an on-device layer. The foundation layer is expected to provide the core model capabilities for Xiaomi’s AI system more broadly, while the cloud layer focuses on engineering work that wraps those capabilities into interfaces used by different products. The largest shift is on-device development, which is being decentralized to teams tied directly to specific hardware businesses such as smartphones and vehicles.
The restructuring aims to reposition Xiao Ai as the underlying “AI capability bedrock” for Xiaomi’s ecosystem rather than a single interaction product managed by one department. Xiaomi partner and president Lu Weibing said that another AI interaction test product called “miclaw,” built on Xiaomi’s MiMo large language model, will not replace Xiao Ai. Instead, miclaw’s capabilities are expected to be integrated into what Xiaomi refers to as “Super Xiao Ai,” described as becoming smarter and more powerful.
A key driver for the architectural change is the shift from a “cloud-centralized” approach, in which devices mainly act as interaction portals calling on cloud processing, toward a “distributed” model where on-device intelligence is developed according to the needs of different terminals. Observers say vehicles are a major factor, given their higher computing requirements, multimodal sensors, and need for low-latency responses that must work closely with local vehicle systems. Under earlier setups, a single centralized cloud roadmap could limit what high-performance devices can do.
With on-device capabilities delegated to operating system teams within individual business units, each line is expected to tailor its AI features to the hardware it ships, such as differences in computing power and sensor configurations. This approach is intended to reduce constraints imposed by a unified product roadmap and better unleash the local AI potential of devices like smartphones and electric vehicles.
Xiaomi’s restructuring also points to a strategic transition in how its AI assistant is expected to evolve. Reports say Xiao Ai’s ecosystem role will support two major directions: more capable long-range agent planning that breaks down and executes complex multi-step tasks, and personalized learning built on a memory system designed to adapt to user habits over time.
