For mortgage broker Peter Liu, AutoCalc started with a problem he was dealing with in his own work: assessing income across different lender requirements could be slow, repetitive and difficult to check.
Payslips vary, and lenders may treat the same type of income differently. Brokers still need to understand what has been captured, check the assumptions and decide whether the result makes sense for the client and lender being considered.
For Liu, the answer was not to remove the broker from the process, but to make the repetitive parts easier while keeping the broker in control of the review.
That idea became more concrete when Sam Zhan joined as co-founder. With senior engineering experience and a background in financial systems, Zhan brought the technical expertise needed to turn Liu’s day-to-day broker frustrations into a working platform.
Together, they faced a broader challenge in financial technology: making professional workflows more efficient without oversimplifying the decisions behind them.
AutoCalc uses AI to organise payslip information and deterministic logic for calculations. Brokers can review the extracted information, reclassify items where required and decide how the result should be used.
The lender’s assessment remains final.
The founders also wanted brokers to see how the platform had handled the information, rather than simply receiving an end result. That visibility matters because even a technically consistent calculation may still require consideration of lender policy, eligibility, suitability and client circumstances.
This approach also shapes how AutoCalc is developed.
Broker feedback is not treated simply as validation or a testimonial. A confusing screen, unfamiliar income scenario or repetitive step may point to part of the workflow that needs to be reconsidered.
Not every suggestion becomes a feature, but frontline experience helps guide what the team tests and improves.
AutoCalc has also taken a measured approach to publishing verification results.
On 4 August 2026, the company compared 166 figures against 27 lender workbooks. Of those, 164 matched, with two differences disclosed publicly on its Accuracy page.
The company describes this as one verification exercise rather than a guarantee of future performance. The 27 workbooks refer only to those included in the test, not the number of lenders supported by the platform.
For Liu and Zhan, the focus is not technology for its own sake, but building a platform around the way brokers actually work.
AutoCalc brings together Liu’s understanding of client and lender requirements with Zhan’s ability to turn those needs into a dependable system.