This chapter provides a comprehensive review of how artificial intelligence is transforming the architecture, operation, and governance of smart energy systems. Drawing on the most recent and influential journal literature, it traces the progression from traditional model-based control to data-driven, learning-based, and physics-informed approaches that now underpin forecasting, optimization, anomaly detection, and predictive maintenance across the electricity grid. The chapter emphasizes that artificial intelligence does not function as a marginal add-on but as an essential technology for enabling resilient, secure, and low-carbon power systems under conditions of high variability, distributed generation, and active consumer participation. Central attention is given to advances in machine learning, deep learning, reinforcement learning, and hybrid models that embed domain knowledge to improve interpretability and data efficiency. In parallel, the discussion examines the rise of digital twins, cloud-to-edge computing, and federated learning as architectural strategies for validating and scaling intelligent applications. It also highlights the most pressing challenges that limit widespread deployment, including data quality and accessibility, interpretability and trustworthiness of models, scalability and real-time responsiveness, and vulnerabilities in cybersecurity and privacy protection. By connecting methodological advances to operational constraints and regulatory requirements, the chapter offers an evidence-based roadmap for translating artificial intelligence from laboratory studies to reliable field practice in modern power systems.