Energy consumption prediction in buildings is an essential element for increasing energy efficiency and achieving sustainability objectives. This paper proposes a conceptual framework based on a formalized conceptual model that describes the relevant variables and the relationships between them in the context of artificial intelligence-based prediction systems. The model integrates input variables such as historical consumption data, meteorological data, building characteristics, and operational data, together with process-related variables including data quality, AI model complexity, system architecture, latency, and data security. These influence system results, expressed through prediction accuracy, operational efficiency, and user trust. The proposed framework highlights the relationships among these variables, emphasizing that system performance depends on the interaction between data, models, and infrastructure. Research hypotheses are formulated to evaluate the impact of heterogeneous data integration, data quality, AI model complexity, and hybrid architectures on system performance. In addition, the role of blockchain mechanisms in increasing trust through ensuring data integrity is highlighted. Unlike existing studies that focus on algorithm performance or technological architectures, this work proposes an integrated conceptual approach, providing a theoretical basis for the design and evaluation of intelligent energy management systems. The model can be used for further experimental validations in real implementation contexts.