AWS described a procedure for modernizing mainframe Assembler to Java using Amazon Bedrock and AWS Transform
AWS described a five-phase methodology for rewriting mainframe Assembler code into Java Spring Boot using Amazon Bedrock, Kiro IDE, and AWS Transform, with COBOL as an intermediate step and human validation of business rules.
Amazon Web Services described a methodology for modernizing mainframe programs written in Assembler, which are tightly coupled with COBOL code, into Java Spring Boot applications running in the AWS cloud. According to the company, the process consists of five phases: extraction of business rules from Assembler code (Business Rules Extraction), creation of detailed functional specifications with human control (human-in-the-loop), conversion of Assembler to COBOL as an intermediate step, functional equivalence testing, and subsequent automated conversion of COBOL code to Java using the AWS Transform tool. If a company already has COBOL source code available, according to the article it is possible to skip directly to AWS Transform.
For the first four phases — rule extraction, specifications, Assembler-to-COBOL conversion, and validation — Amazon Bedrock serves as the inference platform via Kiro IDE (an AI development tool) or the Bedrock console. The company recommends choosing models with strong reasoning over code and a large context window, setting the temperature to 0 for consistent outputs, and for bulk processing of programs using the Converse API with outputs stored in Amazon S3 for the next phase. According to the company, Bedrock does not use prompts or model outputs to train its own models, nor does it share them with third parties.
The intermediate step through COBOL is intended, according to AWS, to serve as a structured, human-readable format that subject matter experts (SMEs) can verify before the final conversion to Java — the goal is to preserve the integrity of the business logic, which, according to the company, a direct Assembler-to-Java conversion does not guarantee with the same reliability. The article emphasizes that AI-generated rule extraction documents need to be treated as drafts that experts must review and validate before any further use.
The remaining parts of the methodology description, including specific prompts and details of the individual phases, are not available in the source text. See the source article for details.
Why it matters
Mainframe programs in Assembler still control critical processes in banks, insurance companies, and transportation, while the number of people able to read them is declining. According to AWS, the proposed procedure with an intermediate step through COBOL and checkpoints is meant to reduce the risk of losing business logic during modernization that today is understood only by departing experts — this is relevant for IT leadership at companies planning such a migration, not for the general public.
Two audiences, two different impacts
What this means
For individuals
Developers working on modernizing legacy mainframe code get a concrete procedure for how to involve AI tools in extracting business rules and converting code, with the recommendation to treat the outputs as proposals requiring human verification.
For a business
Companies operating critical mainframe systems in Assembler (banks, insurance companies, airlines) can, according to AWS, shorten modernization time and reduce the risk of losing business logic thanks to an intermediate step through COBOL with validation checkpoints, instead of a direct rewrite to Java by human effort.
DevelopmentCheck the original
Event sources
only one source so far · 1 publisher, 0 independent. We count feeds from the same owner only once.