A Multi-Layer Framework for AI-Driven Quality Control in Large-Scale Data Production

The integration of artificial intelligence (AI) into quality control (QC) has revolutionized data validation and anomaly detection, enabling automation at scale. However, existing AI-driven QC frameworks often fail to meet operational expectations due to limitations in interpretability, scalability, and false alarm management. In this study, we propose a structured multi-layer framework that addresses these challenges and facilitates the implementation of AI-based QC in large-scale data production. Developed from our experience transitioning the Morningstar Total Return Index (TRI) QC process to an AI-driven approach, the framework improves reliability through scalable data processing, governance, real-time anomaly detection, and intelligent false detection management. By structuring AI-based QC into distinct layers, our framework improves trustworthiness, reduces false positives, and ensures compliance with industry regulations. Validated using Morningstar’s TRI, it demonstrates practical benefits in financial data environments. Through expert feedback integration, dependency-aware anomaly detection, and customizable alerting, the system enhances anomaly detection while minimizing false alarms, making AI-driven QC more adaptive and effective in ensuring data reliability.