The automotive industry is rapidly evolving with technological advancements and increasing customization, leading to a combinatorial explosion of vehicle configurations. Ensuring functionality and quality across these variants poses significant challenges, particularly in balancing rigorous testing with limited resources and tight timelines. This study focuses on the Anlauftauglichkeit Gesamtfahrzeug (ATG) process, a key component of BMW’s quality assurance strategy, ensuring safety, functionality, and compliance before mass production. Traditional testing methods struggle with the scale and complexity of modern configurations. To address this, we propose a novel risk-based testing strategy leveraging industrial data and advanced machine learning. At its core is a graph-based data representation that captures relationships between vehicles, configurations, and test outcomes. This enables the construction of a knowledge graph, which, combined with a Graph Neural Network (GNN)-based framework and Semi-Supervised Learning, facilitates risk assessment and prioritization of high-risk configurations. By integrating domain knowledge and ATG process data, this approach enhances testing efficiency and resource allocation while maintaining high-quality standards. The findings demonstrate the potential of data-driven, machine-learning-enabled methods to transform traditional testing paradigms, with broader implications for other industries managing complex systems under resource constraints.