Finance ops · Production

Lendi Group builds AI-powered Home Loan Guardian using Amazon Bedrock in 16 weeks

The problem

Australian homeowners lacked real-time visibility into whether their home loans remained competitive, and the refinancing process demanded significant manual effort. Brokers spent too much time on administrative tasks rather than high-value client interactions, while Lendi struggled to scale personalized service across their customer base.

Workflow diagram · grounded in source
1
Rate Radar loan monitoring
trigger
“Monitor loan competitiveness by continuously scanning thousands of home loans daily and alerting customers when better deals become available”
2
Initial broker agent engagement
ai_action
“This agent serves as the customer's first point of contact, embodying a friendly, professional persona similar to a human mortgage broker. Its primary goal is to understand the customer's current situation and assess their interest in re…”
3
Customer information collection
ai_action
“this specialized agent systematically collects essential customer details including current loan information, employment status, income, and refinancing preferences. The agent uses conversational techniques to make data collection feel n…”
4
Product recommendation
ai_action
“this agent analyzes the customer's profile against Lendi's extensive database of lenders and products. It presents suitable options with clear explanations of benefits and potential savings.”
5
Application preparation
ai_action
“After the customer selects their preferred product, this agent gathers the additional information required by that specific lender. Different lenders have varying requirements, and this agent adapts its questions accordingly.”
6
Compliance guardrails validation
validation
“Bedrock Guardrails help enforce compliance boundaries, verifying that the customer interactions adhere to Lendi's communication guidelines and remain focused on relevant mortgage-related topics”
7
Linda off-system re-engagement
output
“Linda is the off-system engagement and re-engagement agent that keeps customers connected to their refinance journey, even when they're not actively using the Guardian system. Although the specialized agents manage in-system tasks from i…”
Reported outcome

Guardian has settled millions in home loans with refinance cycle times considerably faster than Lendi Group's baseline, enabling refinancing in only 10 minutes with no paperwork and no phone calls.
The solution was built and launched in only 16 weeks following a more than 30,000-hour cross-functional sprint.

Reported metrics
Home loans settledmillions in home loans
Refinance cycle time vs baselineconsiderably faster than Lendi Group's baseline
Time to complete refinancing10 minutes
Solution build and launch time16 weeks
Show all 6 reported metrics
home loans settledmillions in home loans
refinance cycle time vs baselineconsiderably faster than Lendi Group's baseline
time to complete refinancing10 minutes
solution build and launch time16 weeks
cross-functional sprint effortmore than 30,000-hour
daily home loan scansthousands of home loans daily
Reported stack
Amazon BedrockAmazon Bedrock GuardrailsAmazon EKSAgnoLangfuseMongoDBModel Context ProtocolAurora Digital Twin
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?tag=soumet-20
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Guardian has settled millions in home loans with refinance cycle times considerably faster than Lendi Group's baseline, enabling refinancing in only 10 minutes with no paperwork and no phone calls.

What tools did this team use?

Amazon Bedrock, Amazon Bedrock Guardrails, Amazon EKS, Agno, Langfuse, MongoDB, Model Context Protocol, Aurora Digital Twin.

What results were reported?

Home loans settled: millions in home loans; Refinance cycle time vs baseline: considerably faster than Lendi Group's baseline; Time to complete refinancing: 10 minutes; Solution build and launch time: 16 weeks (source-reported, not independently verified).

How is this finance ops AI workflow structured?

Rate Radar loan monitoring → Initial broker agent engagement → Customer information collection → Product recommendation → Application preparation → Compliance guardrails validation → Linda off-system re-engagement.

Related finance ops cases