How Addepar Scales Investment Workflows with Databricks AI Agents

The lakehouse pattern is finally forcing us to confront a fundamental inefficiency: maintaining separate infrastructure for batch analytics and real-time AI workloads. Addepar's move to consolidate investment workflows on Databricks signals that enterprises are tired of orchestrating data across silos. For data engineering teams, this means your traditional separation of concerns—data warehouse for analysts, feature store for ML, operational database for apps—is becoming a liability, not a best practice. The architectural implication is significant: we're shifting from building pipelines that feed disconnected systems toward building unified data layers where governance, lineage, and access control operate once. If you're still designing around tool proliferation, you're essentially building technical debt. The practical takeaway? Start mapping your existing workflows through a lakehouse lens now. Where could you consolidate? Where does data currently traverse multiple systems unnecessarily? These exercises aren't about jumping platforms—they're about understanding your actual data gravity and whether your current architecture serves your users or just your infrastructure preferences.