Beyond Document Generation: MUFG Integrates Sakana AI for Corporate Credit Approvals
Sakana AI and Mitsubishi UFJ Bank (MUFG) announced that their co-developed "AI Loan Expert" system is advancing to a real-world testing phase following a successful six-month proof of concept (PoC).
The project, a cornerstone of the comprehensive partnership announced by the two firms in 2025, aims to overhaul the bank’s traditional credit approval (ringi) process. Having confirmed the system's operational viability at select pilot locations, MUFG is now preparing for a phased rollout across its nationwide network of branches.
Upgrading the Credit Approval Process
The loan approval process is a critical function in banking, requiring multifaceted evaluations to ensure sound capital allocation. Leveraging Sakana AI’s proprietary agent-building technologies—including its models "The AI Scientist" and "ALE-Agent"—the joint initiative seeks to elevate this process from a manual burden to a highly efficient, AI-assisted operation.
According to the developers, the AI system goes well beyond basic document generation. It is designed to support bank officers throughout the entire corporate lending lifecycle, handling initial data analysis, information organization, financial simulations, and the drafting of final approval documents.
Trial Results and "Human-in-the-Loop" Security
The recent PoC targeted domestic corporate lending and was driven by a joint task force of roughly 100 professionals. The team included MUFG relationship managers, credit officers, and digital strategy staff, working alongside Sakana AI engineers and project managers, with a core group of 30 leading the development.
Operating within a highly secure environment, the AI utilized historical loan data, internal banking regulations, and operational manuals to generate draft proposals. These drafts were then reviewed and finalized by human officers.
The results demonstrated tangible efficiency gains across major lending categories. Participants noted that the system has the potential to serve as a collaborative "partner" capable of supporting both junior staff and seasoned veterans.
Capturing Institutional Knowledge
Looking ahead, MUFG will incrementally expand live, real-case testing across specific branches and departments, with plans to scale the technology to a broader range of lending operations.
Crucially, the companies noted that the project shed light on the importance of capturing the "tacit knowledge"—the nuanced, experience-based intuition of veteran bankers—and revealed concrete methods for embedding this human expertise directly into the AI models.
The Sakana AI and MUFG partnership is not stopping at credit approvals; the two firms confirmed they are actively advancing the AI integration of other banking operations to further sophisticated financial services.
Norbert Gehrke
