Sakana AI: Finance Agents Take Shape

Sakana AI is moving beyond theoretical AI research, launching initiatives to embed its collective intelligence-inspired generative AI into critical sectors like finance. The firm has established an Applied Team to bridge the gap between cutting-edge technology and real-world implementation, with a particular focus on financial services.

The integration of AI agents promises to fundamentally alter how financial institutions operate. To understand this transformation, we spoke with Shota Sakai, a Software Engineer deeply involved in these projects, and Katsuhiro Honda, the Engineering Manager overseeing their efforts. Their work highlights the evolving role of software engineers in bringing advanced AI to market.

Engineers Driving Financial Transformation

Shota Sakai brings a wealth of experience in large-scale financial systems development, having previously worked at Accenture and freee. His background spans core system development, microservices architecture, and product operations. "My career has been centered on large-scale financial systems, expanding into product development and AI," Sakai explained. "At Sakana AI, I'm leveraging that experience to build new AI products, particularly in finance."

Katsuhiro Honda, who joined Sakana AI in August 2025, previously held CTO and Executive Manager roles at legal tech startups and Digital Garage. His expertise lies in leading joint development with major financial institutions and modernizing legacy systems. Honda’s journey began in non-technical roles, but a consistent drive for measurable results led him to engineering, where he now focuses on delivering business value through technology.

Both engineers were drawn to Sakana AI by the rapid evolution of AI agents. "I wanted to confront this change head-on as a developer," Sakai stated. "AI agents have the potential to radically alter software development and business operations, but challenges in reliability, operability, and integration remain."

Honda added, "The emergence of AI coding means the software engineering role is changing. I wanted to be at the forefront of this shift at an AI-centric company." He was particularly attracted by Sakana AI's grounded approach, focusing on substantial market impact rather than hype.

Honda noted a key distinction from his initial expectations: "I initially imagined a research-heavy firm, but I've found customer focus and teamwork to be paramount." The company fosters close collaboration across diverse roles, including Biz, PM, and Applied Research Engineers, emphasizing collective success.

Integrating AI Agents into Banking Workflows

Sakai is currently involved in developing an AI agent designed to support loan origination processes at a bank. This complex workflow involves gathering and analyzing client information, financial data, and business details to draft proposals.

The AI agent aims to assist bank employees by automating initial analysis, information organization, financial simulations, and drafting proposal documents. The goal is not to replace human judgment but to reduce the burden of analytical and documentation tasks, freeing up employees to focus on client interactions and critical decision-making.

Navigating the Technical Landscape

The primary technical challenge lies in establishing quality standards and development processes for AI agents within the stringent constraints of the financial sector. Unlike traditional web applications with predictable inputs and outputs, AI agent behavior can vary based on prompts and context.

Defining "sufficient business quality" for AI outputs is a critical hurdle. Furthermore, financial institutions operate under strict security, cloud, and integration requirements, demanding architectures that are both safe and stable.

This necessitates a holistic design approach, considering UI/UX, evaluation methods, and application responsibilities in tandem with AI agent functionality. Key considerations include managing context, tool execution, permissions, audit logs, human oversight points, and error recovery mechanisms.

Sakana AI's development process embraces AI natively. "Workflows are designed with AI integration from the outset," Sakai observed. "This allows us to leverage AI for efficiency while maintaining rigorous governance and quality assurance through clear instructions, validation environments, and sandbox execution."

Engineers are empowered to focus on higher-level tasks like architecture design, specification judgments, code reviews, and risk assessment. This division of labor, where AI handles repetitive tasks and humans focus on critical oversight, significantly accelerates development velocity.

The team structure comprises Software Engineers, who handle core platform and product development with a broad scope including architecture and developer productivity, and Solution Engineers, responsible for customer integration and addressing specific environmental constraints. The team boasts diverse backgrounds, including web development, data science, SI, cloud, and SRE, fostering a culture of cross-functional collaboration.

Honda expects the team to prioritize successful customer delivery in the short term, emphasizing the ability to navigate real-world operational constraints and collaborate effectively. Long-term, he stresses the importance of feeding insights from individual projects back into the platform and product development cycle, creating a continuous improvement loop.

The Future of AI in Finance

Sakana AI aims to establish its technology and products as the de facto standard for AI utilization in Japan. The vision is to create software where humans and AI collaborate seamlessly within natural workflows.

This involves delivering an integrated experience across UI/UX, application design, security, and operational evaluation. Sakai emphasized, "We aim to co-evolve with our customers, learning from their real-world applications to build solutions that are not only adopted in Japan but can eventually compete globally."

Honda expressed excitement about shaping new development paradigms alongside a globally competitive team. "We're not just using AI tools; we're rethinking organizational structures and processes," he stated, highlighting the push to make AI-driven systems truly viable for enterprise use and to continuously refine platforms based on delivery insights.

Sakana AI is actively recruiting to expand its team, seeking individuals who can contribute to this evolving landscape of AI product development.