"Pageny Will Come Find You" — the Financial AI Concierge, Pageny

Kakaopay unveiled its financial AI concierge 'Pageny', which analyzes a user's location, time, and usage history to suggest optimal financial services and benefits. The core idea is to actively propose services suited to the situation, reducing the hassle of users having to search for services themselves.

Key features include:

• Personalized benefit suggestions: Proactively suggests needed information via speech bubbles by analyzing location and usage patterns
• Multimodal input: Supports both voice (STT) and text input to improve accessibility
• Intent recognition and recommendations: Analyzes hidden intent within complex questions to suggest optimal financial scenarios
• Smart place recommendations: Recommends places and discount benefits by combining user preferences with real-time information

Technically, it adopts a Multi-agent Collaboration architecture built on AWS Bedrock. One Supervisor Agent receives requests and distributes tasks to two Worker Agents (handling structured and unstructured data respectively), generating sophisticated responses through RAG and prompt engineering.

For the architecture, a serverless environment was built using AWS Lambda, and the frontend—implemented in TypeScript and React—was deployed via AWS Amplify. A separate EC2 server is operated for the voice recognition feature to ensure scalability.