Amazon is phasing out outdated software

At this year's in-house trade fair AWS re:Invent in Las Vegas, Amazon Web Services is tackling one of the most persistent problems in corporate IT: technical debt – which inevitably emerges during the aging process of IT infrastructure and software technology as newer techniques replace older ones. And Amazon showcased this effectively: At the fully booked in-house trade fair, which AWS cordoned off for 60,000 visitors, community influencers from the AWS Hero program, as well as selected analysts and journalists, gathered on grounds not far from the conference venue. There, they had a container labeled Tech Debt dropped by a crane onto packages of explosives.

While generative AI has so far primarily been celebrated as an accelerator for creating new software, AWS is now shifting its focus to the refactoring and modernization of old codebases. AWS is massively expanding the capabilities of its Q agent family under the new service, dubbed AWS Transform. The new functions are bundled under this service, but in the background, they use the agent technology that AWS marketed last year under the brand name Amazon Q Developer to intervene deeply in the proprietary logic of companies.

Customized refactoring using natural language

Arguably the most significant innovation is the introduction of AWS Transform Custom. Previous AI tools quickly reached their limits when it came to updating internal, non-public software components. While an AI easily knows the path from Java 8 to Java 17, it has so far lacked the knowledge to replace a company's own authentication library with its successor, for example. AWS is now closing precisely this gap by allowing developers to teach the agent modernization rules tailored to the project in natural language.

Instead of writing complex regular expressions or creating migration scripts, a description of the desired changes is now sufficient, for example, how an outdated internal API should be replaced by a new standard. The AI agent then analyzes the entire codebase of an organization, identifies all affected areas, and independently creates a migration plan. The process runs largely autonomously, as the agent performs the necessary changes across hundreds or thousands of repositories and submits the results as finished pull requests. The role of human developers shifts from the manual drudgery of find-and-replace to reviewing the proposed changes, which promises a drastic reduction in effort for large-scale refactoring projects.

From Jobol to real Java: Mainframe modernization 2.0

AWS has also fundamentally revised its modernization tools under the Transform brand for the classic mainframe sector. A frequent criticism of previous automated migrations from COBOL or PL/I to Java was the quality of the generated code. So-called Jobol often resulted – Java code that is syntactically correct but still looks like COBOL in structure and logic and is therefore difficult to maintain. The new Reimagine functions tackle this with a more intelligent approach, where the AI no longer translates line by line but analyzes the underlying business logic and the intention of the original program.

Based on this, the tool generates new, idiomatic Java code that follows modern architectural patterns and domain-driven design. The goal is a codebase that feels natural to Java developers and no longer carries the legacy of procedural mainframe logic. To minimize the risk of such profound transformations, AWS is also introducing automated test generation. The new tools are capable of recording the functional behavior of mainframe applications and deriving test cases from them that validate the newly generated Java code and ensure functional equivalence.

With these announcements, AWS underscores its strategy to evolve generative AI from a pure assistance system to a proactive tool for managing the entire software lifecycle. In particular, the ability to understand and apply proprietary rules is likely to be a decisive factor for large companies in accelerating the modernization of their established IT landscapes.

Johannes Koch – Principal Engineer Technical Architecture at FICO, co-founder of the AWS Community DACH support association, and present in Vegas as part of the Hero Community – emphasized in an interview with iX: "I can well imagine that in the future, under the AWS Transform brand, AWS will also accommodate more general transformations, for example in the area of application architecture or from one programming language to another. This would make the service an all-purpose tool for simplifying the maintenance or further development of so-called legacy systems."

The pricing of AWS Transform depends on usage – some usage scenarios such as the Mainframe Transformation Agent are free of charge, while custom transformations with AWS Transform Custom are billed by Amazon per agent minute. Exact pricing information can be found at Amazon.

(mho)