Lendi Group just reported 21 per cent year-on-year growth in September-quarter lodgements at its Lendi brand — while ABS data showed new housing loan commitments across the broader market fell 5.4 per cent by number in the June quarter. The difference: two custom-built AI agents named Guardian and Linda that pushed average broker productivity from eight or nine lodgements per month to 13 or 14, without adding a single broker or expanding headcount.
This isn't a story about AI replacing brokers. It's a story about AI doing the tedious work — qualifying leads, collecting documents, chasing missing information — so brokers can focus on the relationship and advice work that actually earns trust. And it's happening at an Australian company, in a regulated industry, with real numbers attached.
What Guardian and Linda actually do
Guardian is the lead intake agent. When a customer enters the Lendi funnel — by text, voice or app — Guardian asks the qualifying questions a broker would normally work through over several calls, then matches responses to live product options in real time. It understands every lender, every lender policy, every state and federal government policy, and can surface relevant information in seconds.
"It takes that customer out of the market, it builds intent, so they're ready to act and transact sooner, and then it puts better qualified customers into our brokers' orbit," said Sebastian Watkins, CEO of Lendi Group.
Linda handles the follow-up. She requests missing information through whichever channel fits the customer's preferences — SMS, email, voice — and re-engages borrowers who have stalled. According to Devesh Maheshwari, Lendi Group's CTO, Linda has driven a more than 20-fold increase in the number of customers who complete the fact-finding stage.
Behind the scenes, it's not just two agents. "There are a swarm of specialised agents behind the scenes," Maheshwari told Computer Weekly. One mimics the group's best customer success associates. Others handle document collection, product-specific information gathering, and refinance monitoring through an AI-powered Rate Radar that scans thousands of home loans daily.
The handoff to a human broker happens when the customer needs advice, has complex circumstances, or faces decisions that materially affect their finances. As Maheshwari put it in the Australian Financial Review: "AI keeps the journey moving; people handle the moments that require judgement and trust."
The numbers that matter
Over a six-month period, Lendi Group reported:
- A 2.5x increase in customers progressing from fact-find to loan lodgement
- Around 20% of customers now complete fact-finding and receive product options, compared with less than 3% through the previous human-led funnel
- 90% of customers who engage with Guardian actively continue the conversation
- Up to 16% of previously disengaged high-intent customers were successfully brought back into the process
"What we're seeing is about the same level of brokers and leads, but what we're seeing is more deals from those leads and more deals from those brokers," Watkins told The Adviser. Processes that historically took weeks can now be completed in under an hour.
Built in-house, not bolted on
What makes Lendi's approach stand out from the growing wave of mortgage AI tools is that the company rebuilt its operations around agents rather than adding AI features to existing workflows.
Maheshwari's team built an in-house orchestration platform called Aurora, starting with the open-source Agno framework. Aurora handles agent configuration through a Mission Control interface (based on Flowgram.ai), stores configurations in MongoDB, and executes through MCP APIs to reach internal and external services. Twilio provides the communications layer connecting voice, SMS and email into a single customer context.
The platform was purpose-built for a regulated industry. Every time an agent is created — even in development — internal agents trigger an alert. Every agent has standard guardrails plus role-specific constraints. Automated testing within production generates alerts when responses fall out of range.
"We don't let models determine the rules. We define the rules because that's our responsibility to our customers to do the right thing. We then allow models to operate within those rules and boundaries," Maheshwari said.
This matters because the broader industry is struggling with exactly this integration gap. A Connective survey of more than 300 Australian brokers found that while 86% believe AI will be essential, 49% say their AI tools are largely disconnected from core systems, and 65% have no documented AI strategy at all. The brokers getting results — like LMG's network reporting 18% faster loan approvals — are the ones integrating AI into end-to-end workflows rather than using ChatGPT in a browser tab alongside their CRM.
What this means for business owners
Lendi Group is not a small business. But the playbook is transferable, and the lesson is specific: the value of AI agents isn't in any single chatbot interaction — it's in connecting every customer touchpoint into a continuous, context-aware workflow.
Three practical takeaways from Lendi's approach:
Start with the handoff, not the automation. Lendi didn't automate the broker out of the process. They automated the gaps between customer intent and broker engagement — the repetitive qualification, the document chasing, the follow-up that falls through cracks. If you're running a service business, identify where customers drop off between initial enquiry and the point where a skilled human adds value. That gap is where agents earn their keep.
Integration beats intelligence. Guardian and Linda aren't impressive because of the underlying model — they're impressive because every channel (voice, SMS, email, app) feeds the same customer context, and every interaction makes the next one smarter. A brilliant AI tool disconnected from your CRM and communication stack will underperform a mediocre one that's properly integrated. We've seen this pattern repeat across Australian enterprises deploying AI agents and financial services firms scaling adoption.
Build guardrails before you build agents. In a regulated industry, Lendi defined the compliance rules first and then let models operate within those boundaries. Whether you're in financial services, property, or professional services, the question isn't whether AI can do the task — it's whether you've defined what AI is and isn't allowed to do. Only 3% of Australian brokers report having formal AI governance policies. That gap is a risk.
What to watch
Lendi Group is now rolling its tested capabilities into Aussie Home Loans, its larger brand. Watkins has set a financial-year-end goal for agents to support the complete customer journey across all scenarios, with brokers able to use as much or as little of the AI-assisted process as they choose.
The broader question is whether Lendi's approach — rebuilding the business around AI rather than adding AI to the business — becomes the template for Australian professional services. Australia still ranks last among comparable nations for AI integration depth, which means cases like this aren't just interesting — they're competitive advantages that compound while competitors deliberate.
At Heygentic, we see Lendi's results as a signal: the businesses that treat AI as an operating system redesign, not a feature addition, are the ones pulling away from the pack. The broker who went from nine lodgements to fourteen didn't learn a new skill. They got a better system around them.
Sources
- Lendi lodgements grow amid downturn as AI rollout accelerates — The Adviser
- Lendi Group Builds AI-Native Home Lending Experience on Twilio — Twilio
- Lendi Group taps agent swarm to reshape mortgage broking — Computer Weekly
- How Lendi revamped the refinance journey using agentic AI with Amazon Bedrock — AWS Machine Learning Blog
- Preparing AI for customers who go off script — Australian Financial Review
- Brokers back AI but most still flying blind on strategy — Australian Broker
- LMG's AI tools cut loan approval times by 18% — Australian Broker
