A new artificial intelligence framework, called NODE-RL-BEM, combines neural ordinary differential equations with reinforcement learning to manage heating, cooling and ventilating in large buildings. Tests reported energy savings of 42% to 48%, comfort violations below 0.5%, and improvements in indoor air quality of 28% to 35%. According to the study published in Scientific Reports, the framework was evaluated using the ASHRAE Great Energy Predictor III dataset and an Intelligent Indoor Environment Dataset, rather than through a live deployment across occupied buildings, meaning the results represent data-driven, real-world-style testing. The researchers report that this approach produced smoother state predictions and adaptive responses, while achieving a generalization score of 0.91 across different operational scenarios.