With the rapid development of artificial intelligence, smart cities require more accurate and real-time sensing capabilities. However, the current sensing systems still face issues such as duplicated construction, data silos, and insufficient intelligence, which hinder efficient urban governance. This study analyzes the challenges of existing sensing systems and explores how AI enhances front-end perception, data fusion, and operational decision loops. It further proposes strategies such as building an integrated sensing network, developing an AI empowerment hub, and strengthening security and maintenance mechanisms to support the optimization of smart city sensing systems.