Implementing AI-Driven Augmented Reality: Insights from an Industrial Use Case Implementation

Abstract Augmented reality (AR) and artificial intelligence (AI) are increasingly applied to improve accuracy and efficiency in industrial workflows. Their combination offers potential for supporting complex manual tasks such as visual quality inspection. Although AR has shown promise in applications like assembly guidance and training, most implementations have been evaluated in controlled environments. Empirical evidence on AI-assisted AR under real production conditions remains limited. This case study presents the implementation of an AI-supported AR system for weld seam inspection in a manufacturing environment. The system integrates CAD-based AR visualization with AI-driven image classification to replace a manual, paper-based inspection process. The approach improved consistency and traceability of inspections, particularly for less experienced operators. Challenges included lighting sensitivity, hardware limitations, and the need for structured image data management to ensure classification reliability. Acceptance among experienced staff was cautious, highlighting the importance of transparent decision logic and user involvement. The results indicate that AI-assisted AR can enhance industrial inspection processes when technical feasibility is paired with human-centered system integration. The approach is transferable to other tasks relying on CAD models and visual criteria and provides practical insights into deploying digital assistance systems in operational contexts.