Dropbox Dash is an AI assistant that connects to 40+ third-party services and ingests millions of resources across every customer's digital workspace. Early on, each integration had its own data structures and storage model. It worked at first, but this approach became a bottleneck: every new integration meant new schemas, new query paths, and new bugs. We rebuilt the system around a generic query engine — a unified data model and search layer that treats all connectors as first-class citizens without requiring connector-specific code at query time. In this talk, you'll learn how we evolved from rigid, vendor-specific storage to a flexible architecture that supports structured filtering and search across all integrations. We'll cover the tradeoffs we navigated, the lessons we learned along the way, and the patterns that let us scale to 40+ connectors without the system collapsing under its own complexity.