Road construction projects are fundamental to infrastructure development but often encounter delays, leading to cost overruns and operational inefficiencies. This study introduces a novel approach utilizing the Generative Pre-trained Transformer 4 (GPT-4) model, augmented with domain knowledge, to predict and categorize the risk factors for delays in road construction projects. Building on prior research, this study extends the models capabilities to include a broader range of delay factors. A thorough literature review was conducted to identify common delay factors in construction. The GPT-4 model, driven by domain-specific insights, constructs a predictive system capable of analyzing the text from project reports, thereby accurately categorizing delays. This study evaluates the model’s performance and demonstrates significant improvements over traditional methods. Notably, the model achieved an 87% F1 score for overall delay risk categorization and a 98% F1 score for project management issues. Additional stakeholder insights highlight the benefits of AI-driven solutions enhanced by domain knowledge for effectively managing delay risks. Future research directions are outlined to refine and optimize the predictive model for better project outcomes.