How NRI Uses Claude on Amazon Bedrock to Automate Japanese Business Document Analysis

Japan faces a structural labor shortage that threatens productivity across its major industries. Nomura Research Institute (NRI), the country's leading consulting and IT solutions firm, recognized that AI-powered document automation could directly address this challenge for its clients.

The target use case was complex Japanese business document review—a task requiring deep comprehension of specialized terminology across financial, manufacturing, and distribution contexts. Standard AI benchmarks, often optimized for English, failed to predict real-world performance on this task.

NRI developed custom evaluations based on actual Japanese business operations and ran head-to-head model comparisons. Claude 3.5 Sonnet via Amazon Bedrock consistently ranked first, excelling specifically on practical client-relevant tasks rather than generic benchmarks.

The deployed solution achieved a 50% reduction in document review times, enabling NRI's clients to automate routine tasks that previously required human reviewers. As Takahiko Inaba, Head of AI at NRI, noted, the system can 'automate routine tasks, enhance productivity of existing workers, and provide 24/7 customer service capabilities—directly supplementing Japan's workforce needs.'