From Banking Code to AI-Powered Speed: How Sony Bank Is Rebuilding Japan’s Core-System Playbook
Japan’s banking industry is entering a new phase of digital transformation, and Sony Bank is emerging as a notable example of how generative artificial intelligence can move beyond customer-facing applications and into the foundations of banking itself. In September 2026, Sony Bank and Fujitsu announced that they had achieved a 30% reduction in the development period for core banking systems, from basic design through integration testing, by embedding generative AI into the development process. The result marks a significant shift in how mission-critical financial technology can be designed, built and tested.
The initiative began in September 2025, when Sony Bank and Fujitsu started applying generative AI to the development of Sony Bank’s core banking environment. By July 2026, the companies had confirmed that development man-hours across the targeted processes had fallen by 40%, while the overall development period had been shortened by 30% compared with conventional methods. Rather than treating AI as a standalone productivity tool, the partners built a system in which AI could work with existing design documents, source code and testing assets throughout the development lifecycle.
The significance of Sony Bank’s experiment lies in its breadth. Generative AI is being applied not simply to writing software code, but to several stages of core-system development, including basic design, detailed design, implementation, unit testing and integration testing. AI agents built around Anthropic’s Claude models and Claude Code, accessed through Amazon Bedrock, are being used across these processes, with human teams retaining responsibility for judgement, approval and quality assurance. This creates a more connected development chain. Design documents can be used to generate source code and test cases, while outputs from one stage can become useful development assets for the next. Instead of repeatedly recreating information as a project moves from planning to implementation and testing, the AI-driven process allows relevant knowledge to flow across stages. Fujitsu and Sony Bank say this approach has helped strengthen collaboration between development processes while improving efficiency.
The figures illustrate the scale of the change. During basic design, the companies recorded up to a 90% reduction in man-hours associated with impact assessments. Detailed design saw up to a 40% reduction in the time required to create design documents. During implementation, the reported source-code generation rate reached 99%, while integration-testing execution man-hours were reduced by as much as 90%. These are process-specific results rather than a claim that every aspect of banking software development has been reduced by the same percentages.
Sony Bank’s AI strategy did not emerge in isolation. It has been supported by a wider transformation of the bank’s technology infrastructure. In May 2025, Sony Bank completed the shift of all its systems to the cloud, including a cloud-native core banking system based on Fujitsu’s Core Banking xBank solution. The architecture runs on Amazon Web Services, creating an environment designed to support scalability and integration with emerging technologies.
That cloud foundation has become important to the generative-AI initiative. Fujitsu’s Core Banking xBank provides a cloud-native environment in which AI services can be integrated and tested more rapidly. For Sony Bank, this means the move towards AI-driven development is being built on an infrastructure already designed for greater flexibility rather than being added to a conventional technology stack as an isolated experiment. The timing is particularly relevant for Japanese financial institutions. Banks operate systems where reliability, security, regulatory requirements and continuity are fundamental. Core banking platforms manage critical functions, making software development fundamentally different from building an ordinary consumer application. The ability to introduce generative AI into this environment therefore depends not only on the capabilities of an AI model but also on governance, data quality, architecture and human oversight.
One of the most important aspects of Sony Bank’s programme is that it has moved beyond a conventional proof of concept. Fujitsu and Sony Bank describe the initiative as an application of generative AI to actual core banking system development, rather than a limited demonstration. AI is being incorporated into live development processes while people remain responsible for final decisions and quality assurance. That distinction could prove important for the wider financial sector. Many organisations have experimented with generative AI for drafting, coding or employee productivity. Applying it to a core banking development lifecycle presents a substantially different challenge because mistakes can have operational and financial consequences. Sony Bank’s approach therefore places the technology within a controlled development framework instead of allowing automated output to operate without human review.
The model also reflects a growing understanding of where generative AI can create value in enterprise technology. The biggest gains may not necessarily come from asking an AI system to write more lines of code. They can come from reducing repetitive analysis, accelerating documentation, connecting information between development stages and automating parts of testing. Sony Bank’s reported results suggest that AI can influence the wider development workflow rather than merely increasing the speed of individual programmers. Japan’s financial institutions have long faced the challenge of maintaining complex technology estates while responding to changing customer expectations. Fujitsu and Sony Bank have noted that the journey from planning a new service to delivering it can take months or even years, while high development costs and complicated processes can restrict the speed at which banks respond to market changes.
Generative AI potentially changes that equation by compressing some of the time-consuming activities surrounding software development. If design information can be converted more quickly into implementation assets, and if testing can be automated more extensively, development teams may be able to devote more attention to architecture, business requirements, security and customer-facing innovation. However, the Sony Bank experience should not be interpreted as evidence that generative AI has removed the complexity of banking technology. The reported 30% reduction relates specifically to the development period from basic design through integration testing, compared with conventional methods. It does not mean that the entire journey of creating and launching a banking service has been reduced by 30%. The distinction is important when assessing what the technology can realistically deliver across the financial industry.
