AI-Assisted Iso-functional Rewriting

Published on:

27 August 2026

Which applications in your portfolio pose the greatest risk, and which ones can wait?

Once the diagnosis is made, you need to take action on the one at the top of the list. AI-assisted iso-functional rewriting is this step.

Why do so many rewrites fail?

Iso-functional rewriting rebuilds the application with identical observable behavior (same screens, same calculations, same rules, including edge cases) while completely rethinking the internal structure. Modernization comes afterward, on a foundation that makes it truly possible because it has been technically mastered.

The role of AI, and why your teams remain central

Freezing a system’s behavior requires knowing what the system does today, and this information doesn’t exist anywhere in written form. An older system contains knowledge that’s missing from its documentation: rules that were never documented, special cases added over the course of updates, and workarounds implemented to address real-world needs on the ground.

This is precisely where AI comes in. It reads the old rather than designing the new. It scans the source code, extracts the rules that are actually being applied, and presents them in a format that your analysts can read, query, and correct. This is the work our solution is already doing across your codebase, applied here to the application you’re rewriting.

Download our White paper

5th-floor-whitepaper-isofunctionnal-rewrite

FAQ

How does this differ from an automated code conversion?

A conversion translates one language into another while preserving the original structure. The result is new but remains difficult to maintain. A rewrite preserves the behavior and rebuilds the structure, which requires architectural decisions made in collaboration with you.

How is this different from a migration or a redesign?

A migration changes the infrastructure without touching the business logic. An iso-functional rewrite changes the technology while preserving the behavior. A redesign changes the behavior—and thus the business process and the associated regulatory framework.

Does AI write the code on its own?

No. It produces a first draft, based on both an analysis of the existing code and code generation. Your teams validate the resulting behavior against the expected behavior. We document this process in detail in our experiment report.

Who owns the resulting code?

You do. This point is contractually locked in from the initial proposal, along with the issue of source code location during the project.

Does this apply only to COBOL?

No. The difficulty stems from the lack of specifications and coupling, not from the language itself. A poorly documented ten-year-old application poses the same problem.

Contributor(s)

Gilles Stragier

Does this sound familiar?

Every organization has its own context, constraints, and pace. Let’s talk about yours. A one-hour conversation is often enough to map out the first leads.

Get in touch

Projects