Future buildings aim to enhance residents’ quality of life while improving safety, security, efficiency, and sustainability. Integrating technologies like Industry 4.0, Internet of Things (IoT), Artificial Intelligence (AI), machine learning, and data analytics offers data-driven solutions for optimizing energy efficiency and predictive maintenance. Despite these advancements, a performance gap persists due to the dynamic nature of smart buildings and the influence of human behaviors and preferences, leading to discrepancies between projected and actual outcomes. Energy management sub-systems in buildings, such as lighting and HVAC systems, often operate inefficiently, especially when misaligned with occupant needs, such as heating empty or partially occupied spaces. Device operation flexibility (direct flexibility) presents significant potential for improving energy management, but this remains heavily reliant on effective information exchange between human occupants (indirect flexibility) and cognitive building (CB) interactive energy management systems. By incorporating hybrid digital twin modeling, advanced predictive and prescriptive analytics can be employed to track device behavior and provide customized feedback, creating a bidirectional information channel that enhances occupant engagement. This approach can improve overall building energy efficiency and reduce operational costs without compromising comfort. This study introduces a novel, user-centric cognitive system for energy efficiency, utilizing demand-side recommendations and thermal comfort controls via multi-agent reinforcement learning. Additionally, a conversational chatbot is explored to facilitate user engagement. The results demonstrate promising outcomes, including a 15.5% reduction in energy load, 69% user engagement, and up to 94% improvement in comfort levels.