Databricks has launched Genie Migrate for Serverless in Beta, an agent that moves jobs and notebooks from classic compute to serverless compute. It scans a fleet, translates compute settings and libraries, and applies changes that can be undone through a 30-day rollback. Each suggested edit arrives as a diff you can check, and nothing gets written until you approve it.
Databricks says thousands of customers already run notebooks and jobs on serverless, and many teams still hold a classic fleet they want to move. The agent works in three steps: it assesses which workloads can move, translates the changes each one needs, and applies them in a reversible way. Changes land on a cloned job first, and you can run that clone once to check it is healthy.
For larger fleets, the agent can run in the background via the Databricks CLI, using the compute-migrations command group to start several jobs from a script. The original job is left alone until you have checked and signed off its move; signing off keeps its ID, schedule, and permissions. Bulk migration covers jobs today, with notebook support coming soon.
Key facts
- Genie Migrate for Serverless is in Beta and runs through the /compute migrate command in Genie.
- Each proposed change appears as a diff that the user reviews before anything is written.
- Uncertain Spark configs are flagged for the user's decision rather than removed silently.
- Applied changes can be reverted with a 30-day rollback.
- Jobs managed by Databricks Asset Bundles or stored in a Git folder cannot be migrated yet.
Why it matters
The agent is available today in Beta. You can call it with /compute migrate inside Genie against any job or notebook. To use it, switch on the beta from the preview options in your workspace. Because it is fully managed, no setup is required before you can begin scanning. Since Databricks runs it, your job config and workload code are read inside your own workspace.




