Databricks bets on declarative DABs
Databricks is trending toward Declarative Automation using DABs (declarative automation blocks) with a push to integrate automated testing into CI/CD and serverless compute for data workflows, according to recent developer posts (x.com). The conversation frames DABs as a way to reduce imperative glue code and automate repetitive testing and deployment steps inside data platforms (x.com).
Databricks has renamed Databricks Asset Bundles as Declarative Automation Bundles and is positioning them as its recommended way to build CI/CD for data and AI projects. (docs.databricks.com) The company’s documentation, updated March 16, 2026, says the bundles package jobs, pipelines, and other Databricks resources as source files so teams can validate, deploy, and run them with the Databricks command-line tool. (docs.databricks.com) Databricks’ CI/CD guide, published April 16, 2026, says Declarative Automation Bundles are now the “recommended approach” for CI/CD on the platform, alongside tools such as GitHub Actions, Azure DevOps, and Jenkins. (docs.databricks.com) In plain terms, the shift moves data teams away from hand-written deployment glue and toward configuration files that describe what should run, where it should run, and how it should be promoted between environments. Databricks says bundles are meant to bring source control, code review, testing, and continuous delivery into data and AI work. (docs.databricks.com) The testing piece is explicit in the new guidance. Databricks’ FAQ says teams should run `databricks bundle validate` in CI/CD before deployment, and its best-practices page says organizations should track test coverage and automate rollbacks for failed releases. (docs.databricks.com 1) (docs.databricks.com 2) Serverless compute is part of the pitch for data workflows. In Databricks’ pipeline tutorial, the company shows a bundle that defines an ETL pipeline and job, then validates, deploys, and runs that pipeline in a workspace on serverless compute. (docs.databricks.com) The examples Databricks publishes now include serverless jobs, data pipelines, dashboards, apps, Unity Catalog objects, and private package workflows, which shows the feature is being used as a packaging layer for more than notebooks alone. Microsoft’s Azure Databricks documentation mirrors the same examples and naming. (docs.databricks.com) (learn.microsoft.com) That marks a broader cleanup of how Databricks wants teams to ship software on its platform. The company’s tutorials describe bundles as a single deployable project, with artifacts, infrastructure settings, and runnable resources stored together in Git-backed files. (docs.databricks.com) There is still a transition underway in the naming. Databricks’ docs repeatedly note that Declarative Automation Bundles were “formerly known as Databricks Asset Bundles,” while community repositories and third-party guides still use the older DAB label. (docs.databricks.com) (github.com) The practical message from the latest docs is straightforward: Databricks wants deployment, testing, and runtime setup described declaratively, then executed through CI/CD and, where possible, serverless infrastructure instead of manual workspace steps. (docs.databricks.com)
