How Thought Machine and AWS Plan to Modernize Banks’ Legacy Cores With AI

Thought Machine and AWS announced an AI-powered approach to legacy core banking migration on September 29, 2026. Instead of translating old mainframe code line by line, the approach aims to identify the banking rules within it and rebuild those rules as Python-based financial products for Thought Machine’s Vault platform. 75

That distinction matters: the proposed automation focuses on recovering what the old system does, then recreating that behavior in a newer core—not simply converting code from one language to another.

How the migration approach works

The companies’ workflow combines AWS Transform with Thought Machine’s Vault Forge:

1. Analyze the legacy application. AWS Transform’s mainframe tools examine and reverse-engineer existing applications to help surface their business logic. AWS describes its mainframe service as covering modernization tasks such as assessment and business-rule extraction. 72
2. Extract the banking rules. The approach focuses on logic behind functions such as interest calculations, fees and loan repayments, rather than treating every line of legacy code as something to translate directly. 5
3. Generate products for Vault. Vault Forge, powered by Amazon Bedrock, uses extracted legacy logic to help create Python-based financial products for Thought Machine’s cloud-native Vault platform. 710
4. Test the generated behavior. Amazon describes the resulting products as simulation-tested; another account of the launch says they can be tested in a sandbox. Testing is part of the described process, not proof that every migration outcome will be correct without further review. 106

Why focus on business rules instead of code translation?

Legacy banking applications can contain years of accumulated product logic. The partnership’s premise is that identifying this logic first may make it easier to recreate specific banking products in a modern platform than to reproduce the old system line by line. The companies say the approach is intended to reduce the time and cost of migration programmes. 75

This is an automation strategy for parts of a complex modernization project—not evidence that every dependency, data migration, operational process or cutover decision is automated. The available announcement material describes rule extraction and product testing, but does not provide a detailed account of bank-team approval gates or the operating model for the joint Core Modernisation Accelerator. 710

Does “years to days” mean a whole bank core can move in days?

No end-to-end timetable for replacing an entire bank’s core is established by the cited announcement materials. They support a more limited conclusion: AI-assisted discovery, rule extraction, product generation and simulation testing are intended to accelerate work that can make modernization lengthy and costly. 710

AWS describes its broader mainframe modernization service as compressing multi-year timelines into months, not days. That is a general AWS service claim, not a demonstrated delivery schedule for this Thought Machine partnership or for a bank’s complete migration. 2

The practical takeaway is that AI may help banks make progress on understanding and rebuilding legacy product logic. Banks should treat speed claims as ambitions for parts of the work, rather than a guaranteed timeline for replacing a live core system.