Idea
Fabric Deployment Pipelines should consistently promote all Prep Data for AI settings with semantic models, including:
- AI Instructions
- Verified Answers
- AI schema and field selections
Today, deploying from Dev to UAT or Production can remove, reset, or inconsistently transfer these configurations. Git/TMDL updates may also cause AI Instructions to return blank.
Why this matters
Without reliable deployment, companies must manually recreate and validate AI configurations in every environment. This:
- Breaks standard CI/CD processes
- Creates inconsistent Copilot responses
- Introduces governance and production risks
- Prevents enterprise-scale adoption
Requested improvements
- Full Prep Data for AI support in Deployment Pipelines
- REST APIs to read and write the configuration
- Reliable Git and TMDL integration
- Deployment validation for missing or changed AI settings
Microsoft Support has indicated that the feature is still in Preview, so support and workarounds are currently limited. Enterprise customers need a scalable, auditable deployment process—not manual configuration in each workspace.
If your organization uses separate Dev, Test, and Production environments, please vote. This capability is essential for making Fabric Copilot enterprise-ready.
