The AI Data Layer: What Agents Read, What the Network Carries, and What Governance Can No Longer See
Enterprises spent two decades building a data stack for humans and reports. Agentic AI is now reading, querying, writing, and recombining that data at machine speed – and the new data layer that feeds it looks nothing like the one that came before. Vector databases, RAG pipelines, embedding stores, and persistent agent memory have become the live substrate of the business. This panel convenes data, security, and network leaders rebuilding the data layer for agentic load: what the stack actually looks like in production, how agentic RAG is reshaping retrieval patterns, the network impact most teams did not see coming (east-west traffic explosion, cross-region sovereignty crossings, new egress cost curves), and the governance gap traditional DLP and DSPM tools are not closing (vector stores outside classification, prompt-assembled contexts that leak across trust boundaries, shadow RAG pipelines built outside sanctioned platforms).
