Sony Bank, Fujitsu cut core banking build time 30% with Claude on Amazon Bedrock

Japan's financial industry has long faced a structural problem: turning a new service idea into a live offering can take months to years, and high development costs plus complex processes make it hard to react quickly to a fast-moving market. Sony Bank and Fujitsu have been working on changing how the core banking system — the foundation of a bank's operations — is developed, together with Amazon Web Services Japan.

The effort began in September 2025, and the companies describe three features that set it apart. Rather than using generative AI in a single step such as coding, they applied it across basic design, build and unit testing, and integration testing, reusing documents, source code, and test assets from one stage in the next. Rather than a proof of concept, the AI was used on a core banking system that is in live operation, across applications handling online transactions and batch processing. And the cloud-native, microservices-based structure of Fujitsu's Core Banking xBank made it possible to apply and evaluate the AI function by function, expanding its scope step by step.

The stakes appear to hinge on whether these results can be repeated beyond Sony Bank. Sony Bank says it aims to become an "AI-native bank" that keeps using AI to respond quickly to changing customer needs, while Fujitsu plans to offer the model to other financial institutions using Core Banking xBank and to reflect the findings in its Uvance for Finance AI Transformation Platform. Whether the same gains hold in those new settings — and in areas like requirements definition, security, and operations and maintenance, where the companies say they will expand next — is likely to determine how widely this approach spreads in Japanese financial system development.