INTRODUCTION
As part of its mission to transform the home loan industry, Lendi Group is working to become one of Australia’s first fully AI-native businesses by 2026. Lendi Group is the award-winning Australian fintech platform behind two of the country’s leading property and home loan brands—Aussie and Lendi—and manages a home loan book of more than AU$107 billion.
Powered by its AI platform, Lendi Group helps Australians to find, buy and own through an integrated ecosystem spanning property search, buyer advocacy, mortgage broking, conveyancing, and ownership tools. The Group supports a national network of approximately 1,350 brokers and 215 retail stores, with a team of around 1,000 people and it’s bringing this vision to life by making AI a core component of its workflows, decision-making, and customer experiences.
To achieve this, Lendi Group has rethought its data architecture. Working with MongoDB, the team moved away from legacy infrastructure and built an operational data layer (ODL), creating the foundation for the next generation of Lendi Group’s AI-powered services. It has already improved time to market for AI features by approximately 40%.
The company’s first launch on the new architecture was Lendi Guardian—a mobile-first, AI-powered home and loan companion, embedded within Lendi Group’s broader end-to-end AI-native architecture. It enables customers to track rates and equity, with one-tap refinancing journeys and pre-filled forms.
According to Lendi Group’s Chief Technology Officer, Devesh Maheshwari, it is like “having an expert watching over your home loan 24x7.” Built on MongoDB Atlas, Lendi Guardian is one step in the Group’s journey to be a fully AI-native business.
THE CHALLENGE
Relational database and microservices sprawl blocking AI-innovation
One of the first challenges Lendi Group took on was to address the fragmented data infrastructure it inherited from the Lendi-Aussie Home Loans merger. Over time, the combined organization's infrastructure had bloated to more than 500 deployable components, built on a mixture of relational, like PostgreSQL, and non-relational databases.
This architecture had become complex, time-consuming, and expensive to maintain. However, that was only half the problem. More importantly, the group’s data infrastructure lacked the consistency and agility required to build and deliver AI services.
“Our developers were spending time tuning and maintaining our databases, rather than pushing features,” said Maheshwari. “To build an architecture that could support our AI-native vision, we needed to reduce our microservices sprawl, reduce complexity, and look at consolidating core operational data into a unified data layer.”
For Lendi Group’s vision to succeed, its AI agents needed a complete, real-time picture of the customer. This meant combining complex, diverse data sets, including property data (e.g., real-time valuations, suburb trends, and geospatial information), finance data (e.g., credit reports and Open Banking feeds), and behavioural data (e.g., customer goals, interactions, and platform usage patterns).
Furthermore, as mortgage broking is tightly regulated, Lendi Group was obligated to meet the highest security and privacy compliance standards. This meant building an AI platform that was compliant from day one and which could easily adapt to future regulations. Doing so would require creating a real-time, immutable audit trail for every single piece of data, and the ability to track who (human, agent, or customer) did what and when. This would ensure transparency and accountability in every decision made by an AI agent.
As the Lendi Group team set out to build a unified operational data layer to support its AI-native strategy, it needed a platform designed to support operational and AI workloads in a more integrated way, and at scale. Handling evolving data structures such as geospatial information, customer interactions, and property imagery with a traditional relational database such as PostgreSQL would mean increasing architectural complexity, slowing delivery, and increasing operational overhead.
OUR SOLUTION
A unified operational data layer powered by MongoDB
Within the first week of the ODL project, Lendi Group had already determined what would power their infrastructure: MongoDB Atlas.
As Will Hargan, Senior AI Systems Engineer at Lendi Group, noted: there simply wasn't another option that offered the flexibility of the document model and the power of MongoDB's integrated, AI-ready data platform. Four key requirements influenced Lendi Group's decision to choose MongoDB Atlas:
Managing complexity
First, the organization needed the ability to manage complex data structures; Lendi Group chose a 'document first' approach. Its aim is to create a unified schema strategy that standardises data contracts across domains. As Will explained, this prevents the frontend and backend from having "two or three different models of the same set of data" and avoids "redefining the shape of the data in multiple places." MongoDB's document model was uniquely suited for this and made it simple for Lendi Group to handle the variety of complex data structures associated with property and mortgage workflows.
AI native features
The second key requirement was finding a platform with AI native features. Lendi Group was building a platform for AI development, not just a data store. MongoDB's flexibility, coupled with built-in AI features like MongoDB Vector Search, enabled Lendi Group to rapidly prototype and iterate on the development of AI applications—all without having to introduce the complexity of a separate vector database. MongoDB Vector Search gave Lendi Group the ability to co-locate the vector embeddings directly alongside the transactional operational data. This simplification was critical for retrieval-augmented generation (RAG) pipelines, ensuring data freshness and consistency while vastly reducing operational overhead.
“A deciding factor in selecting MongoDB was its native support for the data types and workloads required by advanced AI applications,” said Maheshwari. “We knew that integrating AI would require a vector database, and MongoDB’s native vector search capabilities within Atlas help us massively simplify our architecture.”
Scalability
Scalability was the third deciding factor. While Lendi Group already operates at scale, it has ambitious growth plans. MongoDB’s native horizontal sharding enabled Lendi to scale without creating an operational burden and to adapt to the massive data growth anticipated from the future expansion of its AI capabilities.
Security & compliance
The fourth and perhaps most important set of requirements concerned security and compliance. MongoDB provides Lendi Group with critical, built-in security. This enables the company to build an AI business that is secure and compliant by design. Every field in the ODL is tracked with metadata that records who updated it (human, agent, or customer), and when. This continuous audit trail, which is baked into the database layer, ensures Lendi Group maintains complete lineage and accountability, supporting governance controls, traceability, and regulatory compliance requirements.
_Spot.svg)
.svg)