A multinational organization has set up automated AI-driven pipelines to support its customer service operations. After initial deployment, the…

The issues described—inconsistent authentication rules, network routing, and security controls—are fundamental, cross-cutting services that underpin the entire technology stack. This corresponds to the foundation layer, which provides the core IT infrastructure for governance, security, and connectivity. The problem is not with the data, the AI models, or the compute hardware, but with the inconsistent environment foundation upon which these components are deployed. Addressing this layer ensures consistency and stability for scaling AI operations.

A. Data layer: This layer manages data storage and processing. The problem is with access and connectivity, not the data itself.
B. AI/ML platform layer: This layer includes the tools for building and running models. The question states these tools are unchanged and work well in testing.
C. Compute layer: This layer provides processing power (CPU/GPU). The issues are related to networking and security rules, not computational resources.

1. Google Cloud. (2020). AI Adoption Framework. Whitepaper. Page 11, "Part 2: The technology dimension," describes the Foundation layer as encompassing security, networking, identity management, and resource management, which are the core areas of failure in the scenario.

2. Amershi, S., et al. (2019). Software Engineering for Machine Learning: A Case Study. Proceedings of the 41st International Conference on Software Engineering: Software Engineering in Practice (ICSE-SEIP '19). Pages 291–300. https://doi.org/10.1109/ICSE-SEIP.2019.00042. The paper discusses the importance of a reliable infrastructure foundation for deploying ML systems, distinct from the ML platform or data layers.