Databricks is adding an agentic code conversion capability to its Lakebridge toolkit, using Genie Code to provide its new customers another way to move from rival data warehouses to its lakehouse built atop Databricks SQL.
The new capability, called the agentic code converter, uses AI subagents to analyze source code, convert files in parallel, validate the translated SQL, and iteratively retry failed sections within a migration project, while allowing developers to define reusable migration rules for enterprise-specific SQL constructs.
The Lakebridge migration tooling already includes multiple transpilation engines, including the pattern-based BladeBridge technology Databricks acquired in 2025, the compiler-based Morpheus, and the LLM-powered Switch transpiler.
The key difference between the transpilers and the new agentic code converter, according to analysts, lies in how each of them executes conversion and how that could benefit enterprises.
“Deterministic transpilers (Morpheus) handle syntax mapping well: date functions, join syntax, straightforward pattern-to-pattern translation. They get brittle fast on anything requiring control-flow reasoning, cursors, dynamic SQL assembled at runtime, recursive CTEs implemented differently across T-SQL, Teradata, and Oracle,” said Amit Chandak, chief analytics officer at IT consulting firm Kanerika.
“That’s the same 10% to 15% of a legacy codebase that eats most of the manual remediation time on every migration I’ve worked on,” Chandak noted.
In contrast, agentic code conversion, said Pareekh Jain, principal analyst at Pareekh Consulting, is better suited for stored procedures, dynamic SQL, proprietary functions, and complex business logic that require contextual understanding.
“It should improve first-pass conversion quality and reduce manual remediation, especially for the most complex portions of a migration,” Jain noted.
For CIOs, that translates into shorter timelines and lower costs to complete migration projects, especially those that involve thousands of SQL scripts, Jain added.
Chandak, however, said enterprises should temper expectations, mostly because the agentic code converter capability continues to be in beta.
The current beta limits conversion jobs to batches of 300 files, meaning migrations involving thousands of stored procedures still require multiple runs and manual checkpoints, Chandak pointed out.
That limitation could be somewhat offset by the new custom migration skills feature, included with the capability, which allows developers to codify one-off fixes as reusable conversion rules, although whether the agentic system can perform consistently remains an open question till the product matures, Chandak added.
Even if the capability delivers on that promise, Chandak further pointed out that enterprises should not expect it to dramatically shorten end-to-end migration projects because SQL conversion is only one part of the overall modernization effort, as enterprises still need to reconcile migrated data against production systems, sequence dependencies, and obtain business sign-off before workloads can be cut over.
“Because Databricks’ roadmap places data migration and reconciliation in a future phase, the new capability is likely to compress the SQL rewrite effort rather than the overall migration timeline,” Chandak noted.
Databricks isn’t the only data platform vendor using AI to lower the barriers to migration.
Rival Snowflake already has a similar tool, named SnowCovert, that uses agents to help enterprises move their data, data warehouses, business intelligence (BI) reports, and code to Snowflake’s platform.
Google Cloud, too, enables automated migration workflow through Gemini-powered schema and code conversion capabilities in its Database Migration Service.