Modernization needs more than an AI rewrite

Resumo

AI can help rewrite mainframe applications, but generating new code doesn’t prove that an application is ready to run the business. You still need to verify that it processes transactions correctly, meets performance requirements, works with connected systems, and can recover from failures. 

For us, the question isn’t whether AI can produce code. It’s whether the resulting application can meet the business’s requirements and remain maintainable over time. The Gartner October 2026 research report, IBM Mainframe Modernization: AI Claims vs. Realityi, is a great starting point for this discussion.  

We believe replatforming deserves separate consideration from language conversion or rebuilding because the choice changes what teams need to preserve, modify, and validate.

Don’t confuse code conversion with business readiness

In the report, Gartner states: “Modern AI tools remain unproven for automated end-to-end code conversion.”  

We agree. Rewriting an application changes more than its syntax. Teams must establish that the resulting system preserves required behavior, works with surrounding applications and data, and can be maintained over time. 

That’s not a reason to reject AI. It’s a reason to require evidence beyond generated code. Our advice: Use AI where its output can be reviewed and validated, and measure success by tested business behavior and operational readiness.

Replatforming changes the decision

We also believe that when there’s a business reason to move a workload off legacy systems, rewriting the application isn’t the only option. Replatforming enables organizations to retain existing application languages and business logic while moving execution to a supported runtime on distributed or cloud infrastructure. It doesn’t require a wholesale rewrite, though compatibility changes and testing may still be needed. 

Preserving application logic and changing its runtime calls for a different evaluation from converting or rebuilding the application. On the replatforming production record, Gartner states: “Replatform was the first pattern developed for mainframe exit initiatives and remains the only approach demonstrated to succeed for a significant enough number of Gartner clients.”

From Rocket’s perspective, that record is especially relevant when considering Rocket Enterprise Suite. Enterprise Suite is designed to preserve existing application languages and business logic while enabling execution on distributed or cloud infrastructure, giving organizations a replatforming alternative to a wholesale rewrite.  

Start with the workload, not an exit mandate.

Our case for replatforming is selective, not universal. At Rocket Software, we believe platform choice should follow business requirements. Some workloads belong on IBM Z. Others may benefit from a different placement. A move needs a clear advantage, a credible economic case, and a plan to meet operational requirements. 

Start with what’s constraining the business: application change, integration, operating cost, infrastructure choice, or another requirement. Then compare options against the same objectives and time horizon. Neither a new programming language nor a new hosting environment proves that the constraint has been resolved.

Preserve what works and prove what changes

For teams considering replatforming, we recommend answering five practical questions:

1. Which business outcome justifies moving this workload?
2. Which application behaviors, dependencies, and transaction requirements must be preserved?
3. How will testing demonstrate functional equivalence, performance, availability, and recovery?
4. What runtime, infrastructure, skills, and commercial dependencies will the target environment introduce?
5. How will the team continue improving the application after migration?

These questions put evidence ahead of the appeal of an automated rewrite. The last one also changes what you should look for in a platform: support not just for moving and running the application, but for continuing to improve it. That’s the connection we set out to address with Rocket Transaction Platform (RTP).

That need for continued change has shaped our approach to our new platform. We designed RTP to bring replatforming, enterprise transaction processing, and ongoing application modernization together. The goal is to preserve trusted business logic while progressively making application capabilities more modular, testable, and reusable, giving teams a foundation for the changes that follow.

Make the evidence do the work

The same emphasis on tested change guides the AI-assisted capabilities in Rocket Transaction Platform. They help developers understand existing code, document business logic, and generate unit tests while keeping developers in control. For us, AI should support engineering judgment, not replace it.

Our recommendation is to evaluate each approach against the workload’s needs and require evidence that it can deliver. If you’re weighing an AI-led rewrite against replatforming, start with the business outcome you need to achieve. Preserve valuable logic where that makes sense. Prove the operational result.  

Then keep modernizing.