Bean plants are highly vulnerable to various diseases that threaten crop yields and global food security. Accurate and timely disease detection is essential for effective management and the prevention of significant crop losses. Traditional methods, such as manual visual inspection, are inefficient, labor-intensive, and prone to inaccuracies due to human error and subjectivity. This study presents a AI-powered solution that utilizes the complementary strengths of a pre-trained Vision Transformer (ViT) model and a Support Vector Machine (SVM) classifier. The ViT model is employed to extract intricate and high-dimensional features from bean leaf images, providing a rich representation of disease characteristics. These features are then classified using a finely tuned SVM, achieving an outstanding accuracy of 95.00% in disease identification. To enhance interpretability and user trust, Shapley Additive exPlanations (SHAP) is integrated, offering insights into the model’s decision-making process. A comparative evaluation with other classifiers, including NB 78%, RF 94.12%, LR 90.5%, and KNNs 86%, underscores the superior performance of the ViT-SVM combination. This approach not only outperforms traditional methods but also introduces a scalable, interpretable, and efficient solution, paving the way for advancements in agricultural disease management, improved crop protection, and a sustainable global food supply.