AWS Launches AI Agent to Automate Large-Scale Code Modernisation & Reduce Technical Debt

AWS has introduced Transform custom, a new agent to automate large-scale code modernisation and reduce technical debt across enterprises. The company said the service “changes how organisations approach modernisation at scale” by combining pre-built transformations with custom, organisation-specific rules. 

According to AWS, early customers have reported up to 80% reduction in execution time, allowing teams to redeploy developer hours toward product work.

The service is available through the AWS Transform CLI and web console.

Transform custom applies its learned transformation patterns across large codebases, including hundreds or even thousands of repositories. It learns from documentation, natural-language instructions, and code samples, and then improves over time by analysing developer feedback and the manual fixes teams make during the modernisation process.

The service includes a CLI and a web interface. The CLI supports conversational inputs for defining and executing transformations locally or in CI/CD workflows, while the web interface offers campaign-level tracking for modernisation projects across teams.

AWS said the agent supports runtime upgrades for Java, Python and Node.js, and can execute complex transformations such as migrating Spring Boot applications or shifting workloads to AWS Graviton. It can also learn enterprise-specific coding patterns and apply them consistently.

“The service understands not only the mechanical aspects of API changes, but also recognises best practices and optimisation opportunities available in newer SDK versions,” the company said.

Transform custom extends to Infrastructure as Code, with support for CDK-to-Terraform conversions and CloudFormation updates.

A demonstration shared by AWS showed a Python 3.8 Lambda function being upgraded to Python 3.13 using the AWS/python-version-upgrade transformation. The agent analysed the codebase, changed deprecated syntax, updated dependencies, and produced evidence logs. The migrated version was stored in a new branch for developer review.

AWS said that users can iterate as much as needed to refine transformation definitions before publishing them to an internal registry.

Once published, custom transformations can be reused and applied repeatedly across different repositories. The agent generates a tailored JSON plan for each codebase and executes it step by step, supplying detailed evidence for each stage.

AWS said the system centralises modernisation efforts that were previously fragmented across teams. “It keeps institutional knowledge available as scalable assets,” the company said, positioning Transform custom as a way to standardise modernisation while limiting manual rework.