Researching tax lien investments by hand is slow. You pull county records to find delinquent properties, look up each one, check the title, and then decide which liens are worth buying. I wanted to spend my time on the decision instead of the lookups, so I built a tool to do the rest.

What it does

• Pulls the list of tax-delinquent properties from county records.
• Sorts the properties and recommends which tax liens to consider buying.
• Runs basic title searches on the properties I’m interested in.

Background

After the appraisal workflow and agent portal work, MegaLoans turned to the loan application itself: the manual verification that sat between a borrower applying and a processor issuing a pre-approval.

The problem

• Borrowers uploaded pay stubs and bank statements by hand to prove income and assets.
• Processors requested missing documents by hand and created conditions by hand. Conditions are the open items a loan has to clear before it can move forward.
• All of this slowed origination and initial quotes. When income wasn’t verified correctly up front, it came back later as rework and inaccurate quotes.

My role

I owned the strategy and requirements for an AI agent that verifies income, employment, and residency, and I built early prototypes to test the approach. Engineering and data science built and trained the production system.

How the agent works

• It pulls data directly from the source. Income and employment come from payroll providers through Argyle, assets and bank data from Plaid, and identity and residency from Persona. Borrowers no longer have to upload documents for these checks.
• It starts from GSE guidelines. Every file is evaluated against standard GSE guidelines, with our own exception rules layered on top for cases the guidelines don’t settle.
• It has three outcomes. If the agent can clear income, employment, and residency on its own, it issues an automated pre-approval. If it can’t, the file goes to a person in one of two ways: someone checks the agent’s finding, or someone completes that step by hand.

Guardrails

• Monitoring on exception handling, so edge cases get routed to a person instead of failing quietly.
• A feedback button that lets users rate how the agent handled a task. The team used those ratings to track accuracy and decide what to fix first, especially when document formats changed or new edge cases appeared.
• An AI security review before launch.

Results

• Underwriting review time fell 50%, which freed up processor capacity.
• Processing capacity grew to 1,000 loans a month.
• Conversion across the point-of-sale flows rose about 40%, mostly from faster pre-approvals.
• The combination of lower costs and new revenue produced $6M in incremental profit.

What I learned

• Design the handoff before the automation. The most important decision was giving the agent three outcomes. With only pass or fail, every edge case would have landed in one manual queue.
• Feedback keeps an agent working. Without a way for users to flag bad results as they happened, changes in document formats would have gone unnoticed until they showed up in the metrics.
• Leading an AI build takes a different skill. Much of my job was turning vague edge cases into rules the system could act on, and deciding when a person should handle the file instead.

Background

MegaLoans was capped at about 800 rate locks a month. Appraisals took 10 calendar days, and loan officers and processors couldn’t move a file forward until the appraisal came back. Their pipelines stayed full of stalled loans, so they had no room for new clients.

Finding the bottleneck

I talked with operations, underwriting, and secondary to understand where loans were getting stuck. All three groups pointed at the same place. Appraisal speed decided how fast loan officers and processors could clear their pipelines, and that decided how many new loans the company could take.

The goal was 2,000 locks a month. I worked with engineering on a design, sized the revenue the extra volume would bring in, and used that to get leadership to prioritize the work.

What we changed

The old process was built on tasks. People ordered appraisals, chased status, and collected information by hand.

We rebuilt it around data. Each manual task became a data collection step. If we could get the data from another source, the system pulled it and nobody had to do it by hand. Underwriters also got a clear view of the most important thing to work on next, instead of digging through their queues.

Results

• Appraisal turnaround fell from 10 days to 3.
• Monthly capacity grew from 800 rate locks to 2,000.
• With automation and a prioritized queue, underwriters could handle up to 10 times the volume they managed before.

What I learned

The appraisal vendors weren’t the real constraint. The constraint was how much of the process depended on a person doing a task. Once the data came in on its own and underwriters knew what to work on next, the whole pipeline sped up.

Background

Equity Loop is a home equity investment (HEI) company. Homeowners sell a share of their home’s future appreciation in exchange for cash today. Too many of them dropped out during onboarding and the appraisal step.

The problem

• Wait times were long, and appraisal requirements were unclear.
• There was no self-service option, so borrowers contacted support for every status update.
• Manual steps delayed closings and lost deals.

Customer interviews and funnel data in Heap and Google Analytics showed that 80% of borrowers abandoned the process before the appraisal was complete. The main causes:

• No way to see appraisal progress.
• No document upload, so borrowers emailed files.
• No clear list of which documents they would need.

What we built

I led a cross-functional team on a set of self-service changes:

• A redesigned, mobile-friendly borrower portal.
• Live appraisal status pulled from our appraisal vendors.
• Document upload inside the portal.
• A step-by-step checklist that set expectations up front.
• Status lookups through Salesforce and a Glia chatbot, so borrowers could check progress without contacting support.
• A/B tests on the checkout flow.

Results

• 20% more borrowers completed their applications.
• 30% fewer inquiries about appraisal status.
• 7% faster time to funding.
• The 4.9 Trustpilot rating held after launch.

What I learned

• Self-service lowered support costs and let the same team handle more volume.
• Showing live status kept borrowers engaged through the slowest step.
• Small additions like a checklist removed a lot of friction.

The problem

One step in the loan application asked homeowners:

“Do you have any mortgages or liens on your home?”

Many applicants thought of a lien as something negative, like a tax lien, and didn’t realize a mortgage is also a lien. About 20% of applicants who had a mortgage answered the question wrong.

A wrong answer skipped the automated request for mortgage documents. Sales then had to step in by hand, loan estimates were delayed, and the applicant ran into extra back and forth.

What we tested

Test 1: Reword the question. We changed the copy to explain that a mortgage counts as a lien. More people got confused, and errors went up.

Test 2: Change the flow. We dropped the yes/no question and asked for the outstanding mortgage balance instead, with an “I don’t have a mortgage” checkbox. If someone entered a balance, the next page listed lien types and asked them to select any that applied. Incorrect answers fell from 20% to 5%.

Impact

• Fewer applications needed a salesperson to fill in missing mortgage information.
• Loan estimates were more accurate because the mortgage data was there from the start.
• More applications moved through the automated workflow without manual review.

What I learned

• Asking for a number people know worked better than explaining a term they didn’t.
• A small change in how data is collected can remove a lot of manual work later, with no big engineering project.
• Testing beat our first instinct. The obvious fix, clearer wording, made things worse.

I put this together for a product management interview. The prompt was a real problem at WealthTap, a company that offers homeowners a home equity investment (HEI) as an alternative to a HELOC or a refinance.

Homeowners couldn’t tell how long it would take to get an investment offer. That uncertainty hurt customer satisfaction, made sales forecasts unreliable, and had sales and underwriting spending time on deals that would never qualify.

View my full response (PDF)

What was going wrong

• Homeowners had no estimate of when they would hear back.
• Title and appraisal were the slowest and least predictable steps.
• Sales worked deals that later failed underwriting.
• Leadership couldn’t forecast revenue or staffing with confidence.

How I would diagnose it

• Pull historical deal timelines to see where the time actually goes.
• Interview sales, underwriting, and homeowners.
• Map the full workflow to find the handoffs where deals stall.

What I proposed

• Run title and lien searches at the start, so disqualifying issues show up before anyone spends time on the deal.
• Use third-party automated valuations to get an early property value instead of waiting on a full appraisal.
• Add a decision engine that scores risk and recommends investment terms, so unviable deals are filtered out early.
• Show homeowners live status and an estimated timeline in their dashboard.

Expected impact

• Homeowners get a timeline they can plan around.
• Sales spends its time on deals that are likely to close.
• Faster title and valuation decisions shorten the whole process.
• Forecasts improve because fewer deals stall halfway through.

What I took from it

Moving third-party data checks to the front of a workflow is often the cheapest way to make it predictable. Telling customers when to expect an answer also goes a long way, even before you make the answer faster.

Background

MegaLoans lets real estate agents submit refinance and home equity deals for their clients. The agent portal was hard to use, and many agents gave up before submitting a deal.

The problem

• The interface was confusing and required a lot of manual entry.
• Agents couldn’t see where their deals stood, so they contacted support.
• Slow submissions stretched funding cycles and delayed revenue.

Interviews and surveys with agents, plus funnel data in Heap and Google Analytics, showed that 30% of agents abandoned deals before submitting. The main causes:

• A long application with too much manual input.
• No way to track a deal after submitting it.
• No guidance on required documents, which led to incomplete submissions.

What we built

• A redesigned dashboard showing the progress of each deal.
• Forms that pre-fill property and homeowner details from Flueid and CRM data.
• Status updates and notifications.
• Checklists that change with the deal, so agents submit complete applications the first time.
• A/B tests on the interface changes.

Results

• Submission rate rose 30% once the new forms cut down manual entry.
• Agent engagement, measured as time spent in the portal, rose 35%.
• Support tickets fell 40%.
• Repeat submissions from agents rose 25%.
• Cleaner submissions led to faster closings.

What I learned

Agents kept using the portal when it saved them typing and showed them where their deals stood. Giving them that visibility also took a lot of load off support.