The DRG/DIP payment reform requires public hospitals to achieve refined and real-time cost control. However, traditional models face challenges such as crude accounting, lagging control, data fragmentation, and insufficient clinical collaboration. From the theoretical perspectives of cognitive empowerment and decision optimization, this study explores innovative pathways for generative artificial intelligence (AI) to empower hospital cost control. This technology can enable dynamic accounting of disease-specific costs through multimodal data processing, establish a closed-loop, dynamic control system for the entire process based on knowledge reasoning, and promote collaboration between clinical and cost management through natural language interaction. Our analysis indicates that generative AI has the potential to significantly enhance the precision and efficiency of hospital cost control, providing technical support for operational optimization under the payment reform. Simultaneously, the study identifies core challenges including data security, "hallucination" risks, organizational adaptation, and cost-effectiveness, and proposes corresponding strategies. The "cognition-decision" integrated framework constructed in this study aims to guide cost control to extend from end-point control to the optimization of front-end clinical behaviors, offering theoretical reference and practical guidance for generative AI to empower value-based healthcare and facilitate the lean management of clinical pathways.