Lendi’s AI Guardian: Revolutionizing Home Loan Refinancing

Lendi Group, a rapidly growing Australian FinTech, has fundamentally transformed the home loan refinance experience with “Guardian,” an agentic AI-powered application. Built on Amazon Bedrock, Guardian addresses critical challenges faced by homeowners, such as limited visibility into competitive rates and equity positions, and the traditionally cumbersome, time-consuming refinance process. It serves as an around-the-clock companion, continuously monitoring thousands of home loans, providing personalized insights, and significantly simplifying the mortgage journey.

Key features of Guardian include its “Rate Radar,” which scans the market daily and enables refinancing in just 10 minutes with minimal paperwork or phone calls. It tracks equity in real-time and streamlines the application process by auto-populating forms based on internal and external data. The system employs a multi-agent orchestration, featuring specialized agents like the Mortgage Broker Associate for initial engagement, Customer Information Collection, Product Recommendation, and Product-Specific Information Collection. A unique cross-channel “Communication Agent” named Linda proactively re-engages customers via SMS, email, and WhatsApp, ensuring continuous progress throughout their journey.

The target audience includes Australian homeowners seeking competitive mortgage rates and Lendi’s mortgage brokers. Benefits for customers are immense: real-time financial transparency, personalized guidance, and a drastically accelerated, friction-free refinance process. For brokers, Guardian automates routine administrative tasks, allowing them to dedicate their expertise to complex client scenarios, relationship building, and strategic financial advice, thereby enhancing customer loyalty and trust.

Technically, Guardian leverages Amazon Bedrock‘s diverse foundation models and Guardrails for regulatory compliance. It’s deployed at scale on Amazon Elastic Kubernetes Service (Amazon EKS) and uses Agno, an open-source agentic framework, for agent development. Model Context Protocol (MCP) servers integrate institutional knowledge and external services. The architecture features a chat-led UI, an Amazon API Gateway for secure communication, EKS for compute, and an intelligence layer powered by Bedrock FMs. Data storage utilizes MongoDB for user context and conversation history, and Amazon S3 for documents. This robust, AI-first architecture was developed in an impressive 16 weeks, demonstrating rapid development velocity and significant business outcomes for Lendi.

(Source: https://aws.amazon.com/blogs/machine-learning/how-lendi-revamped-the-refinance-journey-for-its-customers-using-agentic-ai-in-12-weeks-using-amazon-bedrock/)