Chiang Mai, Thailand, faces severe air pollution during its annual haze season, driven by agricultural burning, which affects indoor air quality (IAQ) and energy consumption. This study explores the integration of IAQ data, particularly PM2.5 levels, into AI-driven building energy simulations to optimize energy management. Using a custom IoT sensor, real-time data on PM2.5, temperature, humidity, lux, and energy usage were collected from residential settings in Chiang Mai. Feature engineering, including temporal variables and interaction terms (e.g., PM2.5 × temperature), was applied to improve predictive accuracy. Machine learning models, including Random Forest and Gradient Boosting Regressors, were trained to predict energy consumption based on IAQ and environmental conditions. The Gradient Boosting Regressor outperformed Random Forest, capturing complex non-linear relationships between PM2.5 and energy demand. Scenario simulations revealed increased energy usage during high PM2.5 periods, demonstrating the impact of poor air quality on HVAC demand. Conversely, lower PM2.5 conditions showed potential energy savings. These findings highlight the significant role of air quality in energy consumption and emphasize the need for integrating IAQ data into predictive models. This research offers valuable insights for sustainable energy management in Chiang Mai and other regions facing seasonal air pollution challenges.