Objective: To address the dual reform requirements of the New Medicine strategy for healthcare service systems and workforce development, this study explores pathways for AI-empowered medical alliance innovation and systematically constructs compatible training models for primary healthcare professionals, collaboratively addressing the practical challenges of uneven resource distribution and weak primary care service capacity. Methods: Based on the Technology-Organization-Environment (TOE) theory, the research establishes a smart medical alliance theoretical framework integrating technology embedding, organizational collaboration, and institutional support. Within this framework, the study not only designs AI-driven resource sharing mechanisms including expert resource sinking and intelligent equipment scheduling, but also innovatively proposes a closed-loop “training-practice-evaluation-optimization” intelligent training system for primary healthcare professionals, deepening technology application and talent practice through scenarios such as clinical decision support and medical image recognition. Results: Building on this foundation, the research develops an integrated “Smart Medical Alliance+”implementation pathway encompassing “platform + mechanism + talent”, covering modular platform architecture, data standards, and supporting performance evaluation and policy guarantee systems. Conclusion: The findings indicate that AI-empowered medical alliance construction can significantly promote the sinking of quality resources and implementation of hierarchical diagnosis and treatment, while systematic talent training models are the key foundation for ensuring sustained technology effectiveness and achieving equitable, efficient, and sustainable development of primary healthcare service systems.