This work presents a generic, AI-driven framework for the automatic detection, classification, and characterization of surface defects using multi-source image data. Unlike traditional methods, our approach is not confined to specific product types, enabling broader applicability across industrial inspection scenarios. The workflow integrates four key stages: (1) anomaly detection to localize potential defects, (2) precise boundary detection, (3) classification of defect types using fused multi-source imagery, and (4) defect characterization through the integration of spatial features and classification outputs. Standard defect classes, developed through collaborative efforts, are used throughout training and evaluation, ensuring consistent and scalable deployment. This system offers a robust, interpretale, and adaptable solution to modern surface inspection challenges.