As part of our Summer 2026 release of Imogen, our mainframe modernization automation platform, we can now incorporate extracted business rules from AWS Transform directly into Imogen. This follows a similar integration with Google Mainframe Assessment Tool (MAT)  in our Spring 2026 release.

In this post, we explain why customers have asked to use Business Rules Extraction (BRE) analysis from AWS Transform to manage their Imogen modernization programs. We explore what that integration does — and why verification is non-negotiable when it comes to building critical systems from extracted business rules.

How AWS Transform BRE works

AWS Transform Business Rules Extraction (BRE) reads legacy source code and produces a structured, plain-language map of the application, organized from lines of business down through business functions, features, and component-level rules. For planning, scoping, and wave sequencing, this is a real improvement over asking engineers to manually reverse-engineer intent from decades-old code. AWS has also enhanced this in version 2.0 to go beyond the LLM-exclusive analysis that people can do on their own. This provides context for the domains and domain-centric rules, which then need to be implemented at a precise level of fidelity.

‍

The coverage problem

BRE distills the business rules embedded in the code, which is what makes it so useful for planning: it surfaces a representative subset of the conditional logic so teams can scope and sequence the work. But that same selectivity means the implementation details underneath those rules, and the tests needed to confirm they’re fully captured, still require separate enumeration. File organization, numeric encoding, subroutine call semantics, and runtime error handling are part of how the program behaves, but because they're not labeled as "business logic," extraction has no reason to surface them.

‍

‍

Passing tests only matters once you know those tests fully capture the behavior that is important to the system. Coverage is only meaningful when you can measure the completeness of this behavior being modeled; otherwise, you're writing a rubric you know you can pass, but may not really test all the edge cases of your live system. 

When teams feed extracted rules into an LLM to generate modern code (sometimes even the same LLM used to extract the rules), the candidate code never has a chance to run against the live behavior of the program. If you’re trying to build a new system without full coverage of the live behavior down to the implementation details (every branch, edge, and path), you will be building on a system that is internally consistent, but in our experience is highly likely to be wrong about things like encoding, file organization, external call semantics, runtime integration, or critical issues like data type precision. These elements represent the difference between a smooth cutover and one that will take weeks or months to unpick, and in some cases, the difference between rounding transactions up or down.

For more technical detail, read our more detailed whitepaper, Mainframe modernization is a verification problem.

‍

What verification against the running system looks like

That’s where Imogen comes in. It’s built on the premise that the only way to confirm behavioral equivalence is to use the live behavior of the original program as the comprehensive specification. Once a migration plan is in place, Imogen will rapidly recreate an exact behavioral and functional replica in modern code, which makes reimagining far safer and more reliable. We define this behavioral equivalence based on a robust version of coverage that goes beyond naive LLM-generated tests, or simple measures based on covering lines of code.

Coverage is only meaningful when you’re measuring against the live behavior of the system, based on testing the possible paths for execution in the context of the live system. These coverage tests will provide the same input (the same state) to both the original programs and the candidate modern code, and compare the resulting behavior and data outputs for byte-level equivalence. Those inputs are drawn from production-representative data — not AI test cases — so the comparison reflects actual program behavior across the full range of real-world conditions, not just the scenarios an AI or developer might have anticipated. The completeness of these tests is validated by a lineage against the static code, and then also against the live behavior of the system still running on the mainframe.

Any divergence that surfaces is provided back into the code-generation process immediately, regardless of whether it involves a rule that was ever labeled as such. The Imogen platform then uses those tests, code, and a combination of LLMs to automatically implement a modern program that matches the behavior down to the important implementation details like precision and rounding.

With the new integration, the business rules that AWS Transform extracts flow directly into Imogen's overview, providing a clear lineage between rules and the executable verification layer. These rules are mapped to the code implementations that come out of Imogen, giving you a firm footing for your future architecture.

‍

‍

The integration is simple: simply ask questions in natural language referring to context from AWS Transform after configuration in the Imogen user interface.

‍

‍

Unleashing safe reimagination

BRE is valuable for planning, scoping, and giving business and technical teams a common language, and starting to consider what a future architecture should look like. To turn this into a complete implementation specification, you need the tests and rigour that go into a full high-fidelity implementation that you can confirm is behaviorally equivalent.

Organizations that are most effective with BRE output are expecting to deliver on two hopes: first, getting to a new architecture that preserves the outcomes (data) that are still needed by other systems and internal/external stakeholders; and second, considering if these business processes can completely change. The Imogen platform is critical for delivering on this vision, providing a new architecture that becomes the basis for reimagining a system you will be happy to run for the next 20 years.

‍

Learn more about Imogen here, or get in touch to see a full demo.