
Mainframe applications run much of the world’s banking, insurance and transaction processing, and they are notoriously hard to change. On November 17, 2026, Rocket Software announced a new answer: Rocket Transaction Platform, a platform meant to let enterprises replatform, reuse and continuously modernize mission-critical applications on distributed and cloud environments while keeping their business logic and transaction integrity intact.
The launch, described in Rocket’s press release, arrives with an unusually cautious message about AI: it can speed modernization up, but it should not be treated as a shortcut through wholesale rewrites.
What Rocket Transaction Platform is
Rocket describes the product as a highly compatible platform that supports mainframe application semantics. It brings together application reuse, enterprise transaction processing and continuous modernization in one place. It supports application logic in several enterprise languages, including COBOL, PL/I, Java and Python, with support for additional languages such as .NET planned for future releases.
The idea is that teams can progressively expose, extend and evolve trusted application capabilities using modern languages, APIs and development practices, without recreating the proven business logic at the center of their systems.
A portfolio built around choice
Rocket frames the launch as part of a portfolio that does not prescribe a destination. The company lays out three paths:
- Modernize on: customers committed to the mainframe can modernize operations with Rocket EVA, the company’s agentic AI platform.
- Replatform: customers facing skills and agility challenges can move applications with Rocket Transaction Platform while keeping trusted logic and transactional integrity.
- Hybrid: many enterprises will do both, choosing the path workload by workload.
In the release, Phil Buckellew, President of Infrastructure Modernization at Rocket Software, said discussions should start with the business outcomes needed for each application, which often include agility, skills and the economics of running it.
The cautionary note on AI
The most striking part of the announcement is how it uses outside research. Rocket quotes Gartner’s October 2026 research, which states that modern AI tools remain unproven for automated end-to-end code conversion, and that overestimating those capabilities will cause most 2027 mainframe exit projects to fail. The same research, as quoted by Rocket, calls replatforming the only approach demonstrated to succeed for a significant enough number of Gartner clients.
That is a vendor choosing a favorable analyst quote, so it is worth reading as positioning. Still, it reflects a real debate: generative and agentic AI are good at understanding, documenting and testing old code, but converting decades of validated logic automatically is a bigger claim.
What teams can do with it
Rocket lists six capabilities for the platform:
- Preserve trusted applications by reusing them as is on a choice of platforms, including coordination and recovery for transaction processing.
- Expand infrastructure and language choice across distributed and cloud environments.
- Create composable, reusable and API-enabled business services for new applications and AI-enabled workflows.
- Improve development agility with faster build-test-learn cycles.
- Accelerate AI-assisted modernization, such as understanding code, documenting business logic, guided development and generating unit tests, with developers keeping control of changes.
- Enable continuous modernization over time rather than one large transformation.
IDC’s Matthew Flug is quoted in the release saying that relocating an application does not always deliver agility by itself, and that approaches that let organizations modernize workload by workload can provide a more practical foundation.
What it means
For IT leaders, the announcement reinforces a pragmatic view of modernization: keep what works, make it modular, and let AI assist rather than replace human judgment. A platform that runs mainframe-style transaction applications on distributed and cloud infrastructure could widen the pool of developers able to work on them, since Java and Python sit alongside COBOL.
Buyers should still ask hard questions. How compatible is it with a specific application estate? What does migration effort look like? How are performance and recovery behavior verified? The release does not give pricing, availability dates or customer results, so those remain to be seen.
Conclusion
Rocket Transaction Platform is a bet that the safest route to an AI-ready enterprise runs through incremental, workload-by-workload change rather than big-bang rewrites. Whether it delivers will depend on real deployments, but the framing is a useful reminder that modernization is a series of choices, not a single destination.