Sony Bank and Fujitsu confirmed that applying generative AI to the development of Sony Bank's core banking system shortened the development period from basic design through integration testing by 30% and cut man-hours by 40% compared with conventional methods, as confirmed in July 2026. The two companies announced the result on 14 September 2026.
The work began in September 2025. The companies built a mechanism that lets AI use development assets such as design documents, source code and test assets, then expanded its use step by step across the main design, build and test phases of real development. It is applied to the production core banking system, covering banking applications from online transaction processing to batch processing, rather than being limited to a proof of concept.
By phase, impact-assessment man-hours in basic design fell by up to 90%, design-document creation man-hours in detailed design fell by up to 40%, the source code generation rate reached 99% in the implementation phase, and test-execution man-hours in integration testing fell by up to 90%.
The core of the setup is a set of AI agents centred on Anthropic's Claude and Claude Code, using Claude and other models provided on Amazon Bedrock. Agents for design, build and test work across phases on the same assets: source code and test cases are generated from design documents, and the output of one phase is reused in the next. The agents connect to the development assets and tools they need, which the release describes as MCP (Model Context Protocol), Tools and Skills, while developers keep judgement, approval and quality assurance in human hands. Fujitsu staff with financial-operations and system-development expertise evaluate and improve the AI's work at suitable points.
The system runs on Fujitsu's Core Banking xBank, a cloud-native, microservices-based core banking solution operating on Amazon Web Services. The release says this environment made it quick to connect and verify generative AI services, and the microservices structure allowed the AI to be applied and evaluated function by function, so its scope could be widened step by step.
Tatsuya Fukushima, Executive Officer, Sony Bank Incorporated, said: "Sony Bank considers the advancement of system development capabilities as one of its critical management challenges to respond quickly to changing customer needs and business environments. The successful full-scale application of generative AI to the development of our production core banking system is a significant step towards becoming a bank that continuously evolves based on AI."
Sony Bank and Fujitsu plan to extend generative AI to the whole development process, including requirements definition, then to security and to operation and maintenance, and to reuse what they learned in peripheral systems. Sony Bank says it aims to become an "AI-native bank".