Purpose: An AI-driven solution is provided for predicting diabetes risk and managing future glucose levels using advanced machine learning models. By analyzing features such as age, gender, hypertension, BMI, and blood glucose levels, we employ models like Logistic Regression and Random Forest to assess diabetes risk. Methods: We leverage Random Forest model and Long Short-Term Memory (LSTM) networks on CGM data to forecast future glucose levels based on historical readings, insulin doses, and carbohydrate intake. The models are evaluated using performance metrics such as accuracy, loss, confusion matrix, ROC curve, MAE, and RMSE. Results: The best-performing model is integrated into a web application, offering real-time glucose predictions and personalized recommendations. Generative AI further enhances the solution by providing patients with actionable suggestions for improving their health and supporting self-management of diabetes. Conclusion: This innovative tool has the potential to enhance long-term diabetes management by offering data-driven insights and actionable suggestions, particularly benefiting patients who require personalized care to maintain stable glucose levels.