Knowledge graphs become enterprise AI's foundational layer

The shift toward AI-native enterprise architecture reflects a fundamental change in how organizations approach data and decision-making. Rather than treating data as static assets to be queried, enterprises are building systems where AI can reason across multiple sources and explain its answers in context. This requires moving beyond point solutions — a single data warehouse or database — to integrated stacks where different specialized systems (lakehouses for bulk storage, operational databases for transactional queries, time-series stores for temporal data) each play a role. The graph layer acts as the connective tissue, mapping relationships and providing the context that allows models to go beyond simple retrieval and answer follow-up questions like "tell me how you figured it out and give me all the things it's connected to," as Baker phrased it.

However, Baker's candid observation that "unless somebody's withholding information from me, I don't know anybody that's actually totally solved this one yet" signals that this architecture is still emerging and not yet operationalized at scale. The practical obstacles are both technical and organizational. Technically, enterprises must accurately describe their data so AI can interpret it correctly — a challenge Baker frames as a data quality and governance problem ("garbage in, garbage out"). Organizationally, the movement of metadata and access control to a higher semantic layer means that data governance can no longer be confined to database administrators; it must now involve legal and business teams who will specify policies in terms of roles and data categories rather than technical infrastructure details.