UNISOC is driving AI Agents from the demonstration stage into production scenarios and has proposed a "4+1" standard for deploying Agents in production: security first, while balancing stability, observability, and measurability, with human-machine collaboration layered on top. Currently, the company has deployed three types of Agent applications: intelligent investigation of Android Camera memory, sales and supply chain forecasting, and project risk and progress tracking—achieving, respectively, reuse of expert experience, support for business decision-making, and automatic identification of project risks. At the same time, UNISOC has built an internal UniMaaS model service platform that connects full-chain AI capabilities and covers all departments across the company. Going forward, the company will deepen relevant cooperation, enhance the capabilities of the MaaS platform, and advance private deployment and implementation of vertical chip scenarios.