Databricks pushes AutoCDC

Databricks is telling customers to stop writing custom change-data-capture code and let its AutoCDC tools handle database updates automatically. (databricks.com) Change data capture is the plumbing that copies inserts, updates, and deletes from operational systems into analytics tables, and Databricks said those pipelines often break on out-of-order records, duplicate events, and delete handling. In a March 24, 2026 blog post, the company said AutoCDC can automate Slowly Changing Dimension Type 1, Type 2, and snapshot-based patterns with declarative logic instead of hand-built jobs. (databricks.com) Databricks backed the pitch with a customer example from a Fortune 500 aerospace and defense company, where a senior data engineer said four lines of Python replaced about 1,500 lines of prior code. Microsoft’s Azure Databricks documentation says `AUTO CDC... INTO` reads from a streaming change-data-capture source and applies changes to a streaming target table. (databricks.com, learn.microsoft.com) The push landed alongside a wider April product drumbeat around developers building artificial intelligence systems on governed data. Databricks’ April 2026 release notes list a Supervisor API for agents in beta, Databricks Apps navigation changes, governed tags in general availability, and an AI Gateway feature that governs Model Context Protocol servers in beta. (docs.databricks.com, learn.microsoft.com) Adobe is one of the clearest examples of where Databricks wants that data plumbing to end up. On April 20, 2026, Databricks said Adobe Experience Platform can use Delta Sharing for zero-copy access to Databricks data, while Adobe’s documentation says Experience Platform also offers a Databricks source connector for ingestion. (databricks.com, experienceleague.adobe.com) RESAAS made the same argument from a vertical-software angle on April 23, 2026. The real estate software company said its enterprise customers can stream referral activity, transaction pipelines, and live market data directly into Databricks’ Lakehouse Platform. (prnewswire.com) Databricks’ message is that enterprises do not just want new models or chat interfaces; they want fewer brittle pipelines between source systems, governed data stores, and downstream applications. The company’s March and April announcements tie that sales pitch to analytics tables, marketing systems, and industry data feeds that can be shared without copying data into another stack. (databricks.com, databricks.com, prnewswire.com) That leaves AutoCDC less as a standalone feature launch than as a bid to make Databricks the default control layer for how enterprise data changes, moves, and reaches artificial intelligence tools. The company is now pairing that argument with product releases and partner integrations dated March 24, April 20, and April 23, 2026. (databricks.com, databricks.com, prnewswire.com)