As AI technology becomes deeply embedded in organizational operations and decision-making processes, leaders are increasingly faced with new ethical challenges, such as data misuse, algorithmic bias, and unclear accountability. These challenges highlight the urgent need to restructure the corresponding behavioral norms. By systematically reviewing and integrating relevant theories, this paper constructs a leadership framework based on four core mechanisms: Theme Recognition, Dual Rationality, HeXie Coupling, and Dynamic Consistency. First, the Theme Recognition mechanism identifies the value focus that should take precedence when addressing the conflicts between technological innovation and ethical norms. Second, the Dual Rationality mechanism explains how leaders balance tool rationality and value rationality in innovation decisions within the AI context, and addresses the risks of imbalance. Third, HeXie Coupling reveals the synergy and tension between formal systems, technical rules, and ethical norms surrounding the theme and decisions. Fourth, the Dynamic Consistency mechanism emphasizes how organizations can maintain alignment between technological practices and societal ethical expectations through continuous feedback, learning, and adjustment as AI technology evolves. This paper, in essence, aims to address the paradoxical challenges between technological innovation and ethical responsibility in the AI era, offering new theoretical perspectives and institutional frameworks for reconstructing leadership behavior norms, optimizing AI-driven organizational transformation, and enhancing organizational adaptability, innovation, and long-term sustainability, particularly in the Chinese context.