Taku Hiraoka stood before thousands at SoftBank World with a striking statistic: Japanese workers spend up to 2 hours per day creating documents—roughly 30 trillion yen worth of labor annually dedicated to presentation preparation.
"Today, we're here to solve the massive burden of document creation that has held back Japan's white-collar productivity for 25 years," he explained.
Since the 1990s, Japan's white-collar productivity has steadily declined while blue-collar efficiency remained world-class. Hiraoka's theory on why? Japan's consensus-driven culture requires extensive documentation for every decision. SoftBank's internal survey confirmed employees spent hours daily on document creation—and found this stressful.
Executive Lead Hiraoka and Product Lead Kazuki Tajima built Satto Workspace to fix this. The AI-powered platform transforms simple text inputs into complete, professionally formatted presentations in the company's brand and style. Tasks that took hours now take 1-2 minutes.
Building Satto Workspace with Mastra
Technical Lead Yuki Taguchi and his team of 11 engineers chose Mastra as their AI agent framework. "We needed something that could handle complex document workflows while maintaining the precision Japanese business documents require," explains Taguchi. "Mastra's TypeScript-native approach and flexible agent architecture gave us exactly that."
Their technical stack:
- AI Orchestration: Mastra agents for routing between web search, knowledge base retrieval, and document generation
- Frontend: Next.js with custom design system MCP
- Backend: Hono framework with SST for serverless deployment
- Infrastructure: Multi-cloud setup with AWS (primary), Google Cloud and Azure (secondary)
- RAG System: Custom "Satto Cloud" knowledge base integrated with Mastra's retrieval capabilities
- Development: Cursor for AI-assisted coding, GitHub Actions for CI/CD
The platform's "Deep Think" feature particularly showcases Mastra's capabilities—AI agents improve documents through multiple iterations, analyzing and refining content based on company data and user preferences—no human intervention required.
Internal user Hanako Kitano shared her experience: "My boss asked me to come up with a new idea. I was overwhelmed with work, but Satto Workspace created a complete presentation in 1-2 minutes. I presented it as-is and received immediate approval."
Looking Ahead
Using Mastra, SoftBank's team went from requirements to production deployment in just 1.5 months. They're now expanding to mobile workflows, allowing salespeople to update proposals during their commute, and are integrating with Sakana AI's Japanese LLM for better localization.
"We're not just building a product," Hiraoka concluded. "We're reclaiming Japan's lost 25 years and are bringing about a white-collar productivity revolution."
Satto Workspace launches publicly in Spring 2025, with enterprise beta testing currently underway.