Parametric Monte Carlo simulation has become an important technique for quantitative risk analysis in construction projects. However, it relies on experts to manually estimate uncertain activity durations and risks. This paper proposes an artificial intelligence method called AI-SRA to automate schedule risk analysis using machine learning models trained on historical project data. The parametric approach models the project schedule as a critical path method (CPM) network. Experts assign probability distributions to activity durations along with the correlation between risks. Monte Carlo simulation samples from these distributions to forecast the overall project duration distribution. However, this requires extensive manual effort to capture expert judgments of uncertainties. Instead, AI-SRA applies neural network models to predict activity duration distributions directly from schedule attributes. The models are trained on a dataset of hundreds of historical projects with attributes like scope, location, contractors, equipment, crew sizes, productivity metrics, etc. This allows the AI to learn the statistical relationships between schedule features and risk. Monte Carlo simulation is still used, but with activity duration distributions predicted by the AI instead of manual estimation. The AI-SRA method is evaluated on a test set of projects excluded from training data. The resulting project duration distributions are compared to actual outcomes using metrics like prediction intervals. The neural networks are refined until they generate accurate and well-calibrated forecasts. This proposal demonstrates the feasibility of replacing expert estimation with AI for quantitative schedule risk analysis. Automating the extraction of uncertainties from historical data increases objectivity. It also reduces the level of required expertise. The presentation will highlight opportunities to integrate AI more deeply into risk analytics for construction projects.