This study introduces a clear approval process for AI-driven quality control systems to support companies amid unclear regulations. It summarizes applicable regulations, defines stakeholder roles, and provides a nine-phase process based on VDA 5.3 from feasibility to ongoing monitoring, with swimlane diagrams assigning tasks. An interactive cheat sheet maps technical evaluation methods to each approval phase. The result standardizes approval, ensures compliance, and improves transparency while guiding method selection.
| Topic Fields | |
| Published | 2025 |
| Involved Institutes | |
| Project Type | ICNAP Community Study |
| Result Type | |
| Responsibles |
The study defines a structured approval process for AI-driven quality control systems to address unclear regulatory requirements and provide companies with a repeatable, transparent pathway to compliance. Its core functionality centers on a nine-phase workflow, from feasibility analysis through continuous operational monitoring, grounded in the VDA 5.3 guideline for inspection system quality assurance. Each phase is supported by swimlane diagrams that assign tasks and sequences to specific stakeholders, and an interactive cheat sheet that categorizes technical evaluation methods by characteristics and maps them to the relevant approval phase. Architecturally, the approach comprises a regulatory knowledge base, a workflow engine for phase progression and gate reviews, a documentation and evidence repository, and governance controls for role-based responsibilities and auditability. Data flows include ingestion of applicable regulations, mapping to process requirements, capture and versioning of model artifacts and datasets, execution of validation and robustness tests, and ongoing monitoring with alerting and re-approval triggers.
Key technologies and standards referenced include VDA 5.3, and alignment with commonly used frameworks such as ISO 9001 for quality management, ISO 27001 for information security, and emerging provisions of the EU AI Act and data protection requirements such as GDPR. The deployment model supports on-premises or cloud execution, with containerization and API-based integration to QMS, MES, PLM, and MLOps platforms (model registries, data versioning, CI/CD, monitoring). Target users are quality managers, compliance officers, AI engineers, production managers, and supplier representatives.
Performance considerations include timely workflow progression, scalable test execution, and low-latency monitoring pipelines. Security measures comprise RBAC, encryption, audit logs, and controlled handling of sensitive data. Constraints include dependence on accurate documentation, evolving regulations, and domain-specific applicability. Scalability covers multi-site adoption, multi-tenant governance, and phased rollouts. External integrations enable synchronization with document management, ticketing, data lakes, and operational systems to ensure end-to-end traceability and standardized approvals.
Contact us to get in touch! With a membership, you’ll gain full access to all project information and updates.
© Fraunhofer 2025
