This paper presents a hybrid framework that combines a Machine Vision System (MVS) and a Digital Twin (DT), to enable proactive quality control in industrial environments. Grounded in the ISO 23247 standard, the proposed architecture provides support to transform traditional reactive inspection systems into intelligent, collaborative platforms capable of early defect detection and technical factor analysis. For that, the DT integrates a deep learning-based defect identification using the YOLO architecture. Then, a multi-agent platform, developed with SPADE, orchestrates the synchronization between the physical system (MVS) and its DT, supporting feedback mechanisms and enabling data-driven interventions. The methodology is validated through an industrial use case involving the automated inspection of headlamp lenses, a process characterized by multiple defect sources across several production stages. Results demonstrate that the integration of DT and AI technologies significantly improves defect traceability, minimizes rework, and enhances process adaptability, aligning the system with Industry 5.0 principles.