Artificial intelligence is gradually transitioning from being a tool used in the classroom to becoming an organizational capability that affects the way schools interpret evidence, distribute their attention, and coordinate their efforts for improvement. This study looks at how artificial intelligence can help leaders in Saudi secondary schools make data-driven decisions while at the same time keeping professional judgment, equity, and accountability intact. Unlike the reference study mentioned which looks at predictive models for academic outcomes in Saudi higher education, the present paper concentrates on leadership decisions and the organizational factors that make algorithmic evidence useful in the context of secondary schooling. Using a structured integrative review guided by the PRISMA 2020 principles, the review brings together peer-reviewed literature published between 2020 and 2025 as well as relevant Saudi and international policy documents. The evidence is arranged around four questions: which leadership decisions can make use of AI-powered analytics; what capabilities are needed to turn data into action; what ethical and regulatory protections are necessary; and how the introduction of AI can be carried out in the Saudi setting. The synthesis shows that AI is most useful when it aids in the interpretation of early-warning signals, in monitoring instruction, in prioritising resources, in analysing the school climate, and in forecasting operations, rather than taking over leaders' decisions. The benefits depend on the quality of the data, the level of data literacy among staff, the existence of collaborative practices, the explainability of the results, and clear governance. The paper suggests a human-centered decision cycle for secondary schools in Saudi Arabia, connecting the definition of purpose, data stewardship, interpretable analysis, deliberation, action, and monitoring. The review ends by stating that leadership capacity and the level of maturity of governance are more important in ensuring responsible value than the complexity of the algorithms alone.