The product was ready to launch. Then Sundrift AI changed direction

By the time Yarin joined the AI Pathfinder program, Sundrift AI had a product ready to launch. He had already spent about a year developing it. The company had an ambitious technical vision and a clear industry focus. At the point when momentum usually means continuing, however, one question made Yarin reconsider what he had built.

What would stop an established CRM provider from absorbing Sundrift's value into a single new feature?

For a founder who had invested a year in development, that was not a minor product question. It challenged the foundations of the business. Yarin chose to change direction.

The idea for Sundrift began when Yarin took out his first mortgage in Australia. Based in Toowoomba, Queensland, he encountered the lending process firsthand and became interested in the work happening behind it. He began examining how loans move through the Australian mortgage industry and saw an opportunity in the amount of manual work required along the way.

A mortgage must progress from an initial lead through document collection, compliance, assessment, approval and settlement. The relationship can continue long after the money reaches the customer, through refinancing and other changes in the customer's circumstances. Much of the work holding that process together is still administrative.

For Yarin, the opportunity was not simply to make individual tasks faster. It was to create an intelligence and execution layer capable of coordinating the process from “lead to money in the bank,” then supporting the ongoing customer relationship.

Yarin describes Sundrift's central objective as replacing repetitive manual work with processes that are faster, safer, more accurate and less expensive. “Today, processing a deal from A to Z can cost a few hundred, sometimes even thousands, of dollars,” he says. Sundrift is targeting an internal processing cost as low as $5 per deal, an ambitious target the company is now beginning to test through its beta program.

The broader goal is to change where mortgage professionals spend their time. If technology can handle more of the documentation, compliance and workflow coordination, brokers and lenders can concentrate on their clients. Yarin reaches for a deliberately human example: a lender could spend less time managing process and more time understanding a customer's circumstances, following up about refinancing or even remembering an important milestone in their life.

The aim is not only to reduce the cost of lending operations. It is to make room for a better service relationship.

Sundrift AI describes itself as an intelligence and execution layer for the Australian lending industry. At its centre is Vega, the company's model infrastructure. Vega connects with a portfolio of products designed for different parts of the lending market.

One of those products is TOP, which operates as a layer over a customer's existing CRM. It is designed to coordinate lending workflows without requiring a business to replace the systems it already uses. Another product, OS, provides a more complete orchestration environment. Sundrift also offers access to Vega for developers and private lenders building on its infrastructure.

Yarin draws an important distinction between a system that provides information and one that carries out work. Sundrift's agents are intended to coordinate tools, prompts, automations and schedules, then trigger the next action in a process. A prospective customer can interact with Sundrift's digital agent and request contact from the company, and the system can then initiate the follow-up rather than merely suggesting what should happen next.

Sundrift's original product was much heavier. Yarin and his team had built a full CRM with an agentic layer, assuming lending businesses would replace products such as HubSpot or Salesforce with an integrated AI-native system. The technical proposition was compelling. The adoption proposition was much harder.

An established financial business can have decades of customer data, internal processes and automations embedded in its existing CRM. It also has employees responsible for maintaining those systems. Replacing that infrastructure introduces cost, risk and organisational resistance, regardless of how capable the new product may be.

The answer was TOP. Rather than asking a customer to abandon its existing system, Sundrift could sit above it and add an execution layer. “This product came from a need of a client,” he explains. It was a meaningful shift in product strategy: from asking the market to change how it worked to designing around how customers already operated.

Yarin entered the SMEC AI Pathfinder program with the original product ready to launch. Across the program's sessions, the team was pushed to examine the business more deeply. The warning that an incumbent CRM could replicate a standalone feature forced Yarin to confront a question of defensibility as well as customer adoption.

“After a year of developing, I actually changed my mind,” he says. “We pivoted to a different path so we would get a better moat as we proceeded on our journey.”

The program did not identify the mortgage opportunity or build Sundrift's technology. Its contribution was to challenge a critical assumption before the company committed further to the original direction. Almost a year later, Yarin remains convinced that changing course was the right decision.

Yarin says the company owns its model infrastructure and operates it in Australia using Google Cloud infrastructure. Sundrift's current product materials state that customer data remains on Australian servers.

For Yarin, this is not a secondary compliance feature. He sees local infrastructure and control over the underlying technology as part of Sundrift's competitive position. That matters in a highly regulated industry where products may handle financial records, identity documents, compliance information and years of customer data. A lending business needs to understand not only what an AI system can do, but where its information goes and who controls the infrastructure behind it.

Sundrift is betting that Australian ownership, local data residency and technology designed specifically for the Australian lending environment will become more valuable as AI adoption increases.

At the time of his interview, Yarin said Sundrift had early adopters and was preparing to bring a private lender into its beta. The company was also in discussions with a large Australian mortgage franchise.

On 7 July 2026, Sundrift officially opened its beta to selected mortgage brokers and private lenders. The beta represents the next stage of validation: testing whether the platform can fit into the systems lending businesses already use and deliver the operational savings Yarin believes are possible.

The company is still early. Its cost, accuracy and productivity claims will need to be demonstrated through everyday use. But the journey to beta already contains an important lesson about founder progress. Yarin's first mortgage showed him the opportunity. Customer conversations revealed the cost of asking established businesses to replace their systems. AI Pathfinder forced him to confront whether the original product could be easily replicated by the incumbents around it.

Sometimes startup progress looks like launching faster. In Sundrift's case, it meant having the confidence to stop, rethink the architecture and build something designed to be both easier for customers to adopt and harder for competitors to replace.