Saudi Arabia's Vision 2030 positions economic diversification and knowledge-intensive industrial upgrading as national imperatives; however, the mechanisms through which refinery digitalization can plausibly contribute to these objectives remain theoretically underdeveloped. This study develops a multi-level conceptual framework explaining how artificial intelligence (AI) adoption and knowledge management (KM) capability co-evolve in Saudi Arabian refineries and how this co-evolution generates transformation outcomes aligned with Vision 2030 diversification goals. Integrating the knowledge-based view, dynamic capabilities theory, and technology–organization–environment (TOE) framework, the model positions KM capability maturity as the central mediating mechanism through which AI adoption translates into operational excellence, sustainability-oriented innovation, and cross-site scaling outcomes. Trust in AI and sustainability orientation are theorized as boundary condition moderators. Six testable propositions and a three-phase empirical research agenda are advanced. The framework contributes an integrative, refinery-specific conceptual model that bridges Saudi institutional evidence, hydrocarbon-sector AI–KM research, and digital refinery scholarship.