


AI voice agents
Lendi Group · 2024
Senior Product Designer, AI Disrupt
Designing how Lendi and Aussie sound, listen and hand back to people
Voice agents that answer Lendi and Aussie customers after hours and follow up the people associates couldn’t reach. I led the original experience from first call to production: what the agent says and knows, how it recovers, and how it hands over to a person.

About Lendi Group
Lendi Group is a leading Australian property fintech behind the Lendi and Aussie brands, formed through their merger in May 2021. Its 1,300 brokers write around 6% of the country’s home loans every week, and it’s growing into a full property ecosystem that helps Australians find, buy and own. Underneath, it’s rebuilding its platform on cloud-native infrastructure run by AI agents, to turn the complexity of a mortgage into a 10-minute digital experience.
- Year
- 2024
Voice agents for after-hours help and customer follow-up
When Lendi and Aussie closed for the night, customers who called had nowhere to go. And during the day, sales associates spent hours calling people they couldn’t reach. When a call went unanswered and no voicemail was left, the chance of a callback went with it.
Lendi Group’s AI Disrupt team, a product manager, a tech lead and me, set out to fix both with AI voice agents: one to answer people after hours, and one to follow up customers, understand what they wanted and line up the next step. Other teams have since scaled the work, and the voice agents have saved thousands of hours since the pilot.
My part
I led the original voice experience, from the very first call through to production: what the agent could say, what it could know, how the conversation moved, and when to hand over to a person. AI voice was still dominated by American accents, so we cloned team members’ voices to make the agents sound genuinely Australian. We worked with Bland AI, relationship managers, Journeys and compliance.
In testing, I checked the recording, the transcript and the agent’s response separately, so each fix targeted the real problem. When the service went live, I handed it to the operational team with test scenarios, maintenance guidance and open issues, and moved on to Aurora.
In the news
The full case study
The full story, with the research, designs, decisions and results behind it, includes internal work, so it’s password protected. If you’ve been sent a password, enter it below, or email me for access.
