Talking to All Your Data Starts Feeling More Practical

Genie and Lakehouse Federation make it easier to ask questions across data that still lives in different systems

Created by Author

In many teams, data is not sitting in one clean place. Some of it lives in Databricks. Some of it stays in Snowflake, BigQuery, Oracle, Postgres, AWS Glue, or other systems. The business still wants answers across all of it, but getting there is usually not simple.

That is where things often slow down.

A full migration can take time. Copying data creates more work. Building extra pipelines adds more maintenance. Even when the goal is clear, the path can feel heavier than it should.

That is why a recent Databricks update stood out to me. In a June 12, 2026 post, Databricks explained how Genie can now work with Lakehouse Federation so users can ask questions across data that still lives in external systems. Databricks says this lets teams connect Genie to federated sources, bring them under Unity Catalog governance, and start asking natural language questions without waiting for a big migration first.

What I like about this direction is how practical it feels. In real projects, many teams are not blocked because they lack data. They are blocked because the data is spread across too many places. Business users still want one answer, not a long explanation of where the source sits. If the platform can help them ask across systems more naturally, that becomes very useful.

Databricks also highlights something important here. It is not only about connecting sources. It is also about carrying context forward. The post explains that table comments and column descriptions from supported source systems can be federated into Unity Catalog, which helps Genie understand the data better instead of starting from raw table names alone. That is a meaningful step, because AI works much better when the data comes with business meaning, not just structure.

This is where the update becomes more than a simple connectivity story. Databricks says teams can define reusable business semantics and metrics on top of federated data, so Genie, dashboards, and notebooks can use the same trusted logic even when the data has not been moved into Databricks yet. That helps reduce one of the biggest problems in analytics: different systems answering the same question in slightly different ways.

For data engineers, this matters because a lot of effort goes into the space between systems. If users can start asking questions across the estate without forcing a full migration first, teams get more flexibility. They can modernize in stages instead of trying to solve everything in one big move. Databricks even positions this as a fast on-ramp, with the option to improve performance later by converting foreign tables into Unity Catalog managed tables when the time is right.

The biggest takeaway for me is simple. Good platforms should not make teams choose between asking better questions now and migrating everything first.

In modern data work, access becomes much more valuable when it reaches across the real data estate, not only the ideal one.

Have you seen this in your own projects too? Is your team still waiting on migration before users can ask better questions, or are you finding ways to bring insight across systems earlier?