How bunq handles 97% of support with Amazon Bedrock

This article describes how bunq, Europe's second-largest neobank, uses Amazon Bedrock and a multi-agent AI architecture to automate customer support at scale.

• Finn AI assistant handles 97% of bunq support, with 82% fully automated
• Average response time reduced to 47 seconds across all interactions
• Orchestrator pattern replaced router-based architecture to avoid scalability bottlenecks
• Primary agents dynamically invoke specialized agents as tools when needed
• Supports 38 languages with real-time speech-to-speech translation capabilities
• Built on Amazon Bedrock, ECS, DynamoDB, and OpenSearch Serverless
• Deployed from concept to production in 3 months with 80-person team
• Enables image recognition for receipts and intelligent financial insights

bunq's multi-agent orchestrator pattern with Amazon Bedrock demonstrates how flexible AI architecture can transform banking operations while maintaining security and compliance at scale.