From 15 to 28 Loans a Month: A Broker's AI Story

The Challenge

An independent mortgage broker in Western Sydney. Good reputation. Strong relationships with lenders. A solid book of business.

Processing [PLACEHOLDER: 15 loans per month].

That was the ceiling. Not because demand was capped. The pipeline was deeper than that. It was the ceiling because one person couldn't process more.

Here's how it worked: A client would approach with a loan application. The broker would request documentation—bank statements, payslips, tax returns, asset declarations. The client would send documents, sometimes organized, sometimes as a chaotic pile of photos and PDFs.

The broker would review the documents, reconcile them with the loan application, flag issues, request clarifications. Then compile a complete application package and submit it to lenders. Follow up as lenders requested more information. Manage counter-offers. Coordinate settlement.

It was a one-person operation. The broker did everything.

Which meant the broker couldn't take on more than [PLACEHOLDER: 15] loans per month without burning out. And brokers who burn out either quit or stop taking on new clients.

The demand was there. The lender relationships were strong. The only constraint was: how much can one person actually process?

"I'm leaving money on the table," the broker said simply. "I can see the opportunity. I just can't manually process it."

Hiring someone would help, but at [PLACEHOLDER] per application in processing costs, it was a marginal economics problem. Scale the business [PLACEHOLDER]%, and suddenly a new hire makes sense. But get stuck at [PLACEHOLDER: 15-17]] loans a month, and there's no budget for it.

This is the trap most service-based businesses hit: Scale is constrained by manual capacity. Hiring is risky at small scale. So you stay stuck.

The Discovery

We walked through the loan processing workflow in detail.

Document Collection: The broker requests documents. Clients send them in over email, SMS, sometimes through a portal. Some send complete packages. Some send piecemeal. Document review involves:

Checking that all required documents are present
Verifying document dates align with application dates
Flagging missing or outdated documents
Requesting clarifications and re-submissions

This phase took [PLACEHOLDER] minutes per application on average, but varied wildly—[PLACEHOLDER] minutes if the client had everything organized, [PLACEHOLDER] minutes if documentation was chaotic.

Information Extraction: Data had to be pulled from documents and organized in a standard format:

Income information from tax returns and payslips
Asset information from bank statements and investment reports
Liability information from credit reports and loan statements
Property valuation information from valuations or online tools

This was systematic, repeatable work. Extract income amount, extract employment status, extract date of document, flag discrepancies. The broker did this manually—writing notes in a spreadsheet, reviewing documents, extracting numbers.

[PLACEHOLDER] minutes per application, often longer if documents were unclear or information was scattered.

Lender Matching: Once the application was complete, it needed to be matched to the right lender. Which lenders would approve this application? What terms would they offer? What conditions would they impose?

The broker maintained relationships with [PLACEHOLDER] lenders, each with different requirements, pricing, and service levels. Matching an application to lenders required knowing the lenders' appetite—for this income level, this property type, this loan-to-value ratio, this credit profile.

The broker had this knowledge in their head. From experience. From relationship management. From seeing what worked and what didn't.

[PLACEHOLDER] minutes per application to think through the right lenders and build proposals.

Application Compilation & Submission: Preparing the application for lender submission:

Ensuring all documents are in the right format
Creating a cover letter or application summary
Submitting to multiple lenders
Following up on status and required information

[PLACEHOLDER] minutes per application.

Ongoing Coordination: Following up with lenders, managing counter-offers, addressing conditions, coordinating settlement. This was relationship and communication work. [PLACEHOLDER] minutes per application on average, but highly variable.

When we summed it up: [PLACEHOLDER] minutes per application to process start-to-finish. A working month had [PLACEHOLDER] hours available. At [PLACEHOLDER] minutes per application, that was roughly [PLACEHOLDER] applications per month.

The broker was at [PLACEHOLDER: 15]], which meant they were at about [PLACEHOLDER]% capacity utilization. They weren't lazy. They were just doing everything.

The Solution

We built an AI system focused on the tasks that were both repeatable and time-consuming: document collection, information extraction, and lender matching.

Smart Document Collection: Instead of email and chaos, we created a structured document collection system. Clients received a checklist of documents needed and a simple upload interface. The system tracked which documents had been received, which were still outstanding, and automatically sent reminders.

When documents were received, the system validated them: Is this document type valid? Is the date recent enough? Are there obvious issues (blank pages, wrong document, illegible scans)?

Problematic documents were flagged for the broker to review. Valid documents were automatically ingested into the application file.

This reduced document collection time from [PLACEHOLDER] minutes to [PLACEHOLDER] minutes per application—the broker no longer chased clients for documents; the system did that automatically.

Intelligent Information Extraction: Once documents were validated, the AI extracted key information:

Income figures from payslips and tax returns
Employment verification
Asset balances and types from bank statements
Liability information from credit statements
Property details

The system didn't guess. When it was uncertain, it flagged items for human review. But for standard documents—recent payslips, standard tax returns, clear bank statements—extraction was automatic.

This reduced information extraction time from [PLACEHOLDER] minutes to [PLACEHOLDER] minutes per application. The broker reviewed the extracted information for accuracy but didn't have to hunt through documents to find numbers.

Lender Matching Engine: This was the most sophisticated piece. We trained the system on the broker's lender relationships—which lenders they worked with, what each lender's requirements and pricing was, what applications had been successful or unsuccessful.

The system learned the broker's knowledge: This lender prefers investment properties over owner-occupier. This lender has strong service levels but lower interest rates. This lender will approve this credit profile readily. This lender won't touch this type of property.

When an application was ready, the system recommended the optimal three to five lenders—the ones most likely to approve at competitive terms. The broker still made the final decision, but they weren't thinking through the entire lender landscape manually. The system had done the pattern-matching.

Lender matching time dropped from [PLACEHOLDER] minutes to [PLACEHOLDER] minutes per application.

Application Compilation Automation: Once lenders were selected, the system auto-compiled applications—formatting documents, creating cover letters, preparing submission packages.

[PLACEHOLDER] minutes of manual work became [PLACEHOLDER] minutes of review and submission.

All of this ran on a simple, local system. Low cost. Fast implementation. Integrated with the broker's existing tools.

The Results

Processing time per application dropped from [PLACEHOLDER] minutes to [PLACEHOLDER] minutes. That's a [PLACEHOLDER]% reduction.

With the same [PLACEHOLDER] hours per month available, the broker could now process [PLACEHOLDER] applications monthly instead of [PLACEHOLDER].

That's an 87% volume increase without hiring. Without adding overhead. Without renting more office space or changing the team structure.

Results:

[PLACEHOLDER: 28 loans per month]] instead of 15, using the same labor
Same income per application, higher total revenue due to volume increase
Broker no longer bottlenecked by manual capacity
Quality improved slightly—the system didn't miss documents or misextract data; the broker could focus on the relationship and matching aspects instead of data hygiene
Headroom to grow without the next hire being a necessity

The broker: "I went from 'I can't scale' to 'I can scale without hiring.' That changes the business fundamentally."

What's Next

The broker is now exploring ways to expand the system. Currently it covers processing. They're testing whether the same matching logic could apply to refinance applications—a different product with different lender requirements.

They're also considering whether they could use the system to help clients understand their borrowing capacity before they apply—essentially, a pre-application advisor tool that would screen clients against the lender data and show them which lenders were likely to approve their application at what terms.

"The system gave me capacity," the broker said. "Now I'm thinking about what I do with that capacity. More volume, yes. But also, different kinds of value."

The Lesson

If you're running a service business and you're at a processing capacity ceiling because everything requires your manual effort, you have an AI opportunity.

The document collection, data extraction, and pattern-matching aspects of your work are probably doing two things:

Taking up enormous amounts of your time
Constraining your growth

AI won't replace your expertise or your relationships. But it will handle the repeatable aspects of your workflow, freeing your brain for actual client service and business strategy.

Most service businesses think they need to hire to scale. Sometimes that's true. But often, what they actually need is to automate the repeatable work first, then see if hiring is still necessary.

Ready to find the repeatable work that's blocking your growth? Book a free discovery call with CORSZA. We'll help you see where you're manually processing volume—and then we'll give you back the capacity.