Latest / The Operations Podcast with Fexingo: Process, People, and Profit in Modern Business / How a Mortgage Lender Cut Approval Time by 87 Percent
Transcript
- Lucas: Last week, a major US mortgage lender reported something that would have sounded like science fiction five years ago: the average time from application to approval dropped from thirty days to just four. And this wasn't some tiny fintech — we're talking about a bank that originates over fifty billion dollars in home loans every year. Luna: Four days? That's faster than I can get a credit card approved. What did they actually change? Lucas: It's a complete rewrite of their origination workflow. The old process was classic batch and queue: a loan officer collects documents, sends them to a processor who checks for completeness, then to an underwriter who reviews credit and income, then back to the borrower for any missing items — and that loop could repeat three or four times. Each handoff added days. Luna: Right, and this lender was dealing with about fifteen thousand applications a month. So even small delays compound. Lucas: Exactly. The first change they made was to digitize the document intake. Instead of borrowers faxing or mailing pay stubs and bank statements, they switched to a secure upload portal that uses optical character recognition and machine learning to extract key data fields automatically. That cut the initial processing step from an average of two days to about twenty minutes. Luna: Okay, but underwriting — that's the real bottleneck. You can't automate judgment calls, right? Lucas: That's what everyone assumes. But they took a triage approach. They built a model that scores each application for complexity — about forty percent of their loans are straightforward: W2 employees with solid credit, a down payment of twenty percent or more, and a property in a non-complex market. Those get routed to a fully automated underwriting engine that checks against agency guidelines and spits out a conditional approval in minutes. Luna: So for the straightforward cases, it's basically fully automated. What about the other sixty percent? Lucas: Those go to a team of human underwriters, but here's the key operational change: they redesigned the work from a sequential assembly line into a pod structure. Instead of an application moving from person to person, each pod has a processor, an underwriter, and a funding specialist who work together on the same group of files in real time. They use a shared digital workspace so any member of the pod can see the status instantly. Luna: That reduces the wait time between handoffs. But I'm guessing it required a lot of cross-training — and probably some resistance from people used to their siloed roles. Lucas: You're spot on. The operations team spent six months retraining staff and redesigning incentive structures. Instead of measuring individual output — how many files per person per day — they started measuring pod-level cycle time and customer satisfaction. The funny thing is, once the pods got comfortable, their throughput actually increased by thirty percent. Luna: That's the kind of efficiency gain that makes CFOs smile. But let me ask a skeptical question: did the faster approval lead to higher default rates? Did they sacrifice risk assessment for speed? Lucas: Great question. The lender was worried about that too. They tracked it for a full year — default rates actually dropped by fifteen basis points. The reason is that the automated triage doesn't skip steps; it enforces consistent criteria. The human underwriters in the pods can focus on the complex cases where judgment actually adds value, rather than spending time on routine verifications. So the overall quality of underwriting improved. Luna: So the operations redesign didn't just cut time — it also made the process more accurate. That's the holy grail. Lucas: It really is. And the results cascaded through the business. Customer satisfaction scores went up because borrowers weren't left in the dark for weeks. The lender's pull-through rate — the percentage of applications that actually close — increased by eight percentage points, because people didn't get frustrated and switch to another lender mid-process. Luna: That makes sense. In a rising-rate environment, speed is a competitive advantage. Borrowers want certainty quickly. Lucas: Exactly. And this isn't just about mortgages. The same principles — digitize intake, triage by complexity, reorganize into cross-functional teams — apply to any service process that involves review and approval. Think insurance claims, small business lending, even visa processing. Anywhere there's a sequential queue with manual handoffs, there's an opportunity. Luna: And it doesn't require a massive technology investment. The core changes were process and people, not some AI silver bullet. Lucas: Right. The technology was important, but the real heavy lifting was rethinking the workflow and aligning incentives. That's the kind of operations lesson that sticks. And if these conversations have made you think differently about a process in your own work, that's exactly what we're here for — listener support is what keeps this show ad-free and focused on real cases. If you ever want to chip in, it's buy me a coffee dot com slash fexingo. No pressure, but it helps us keep digging into stories like this. Luna: Yeah, and we appreciate everyone who listens. Speaking of digging deeper — let's talk about the exceptions. What happens when an application falls outside the automated guidelines? Lucas: That's where the pod structure really shines. The automated system flags any application that doesn't fit standard parameters — self-employed borrowers, properties in certain condominium complexes, jumbo loans above the conforming limit. Those go straight to a pod that specializes in non-standard files. They've got a dedicated escalation path and a library of pre-approved alternative documentation requirements. Luna: So instead of getting stuck in a generic queue, exceptions are routed to the right expertise immediately. Lucas: Exactly. And the pods have authority to make decisions within certain bounds — they don't have to go up three levels of management for every variation. That autonomy was a big culture shift, but it reduced the average decision time for complex files from ten days to three. Luna: Now, I've heard that some lenders are using AI to generate the initial loan estimate and closing documents automatically. Did this lender go that far? Lucas: They did, but only after they had the core process stabilized. They rolled out a document generation module that pulls data from the underwriting system and fills in the standard forms — the Loan Estimate, the Closing Disclosure. That eliminated another two days of manual data entry and reduced errors. The compliance team caught fewer mistakes because the system enforces regulatory formatting. Luna: It's amazing how much of mortgage processing is just moving data from one form to another. Automating that seems like low-hanging fruit. Lucas: It is, but you have to build the infrastructure for it. The data has to be clean and standardized on the front end. That's why the digitized intake was step one. You can't automate garbage. Luna: Fair point. So what's the biggest lesson for other industries from this case? Lucas: I think it's that speed and quality aren't a trade-off when you redesign the process from first principles. This lender didn't tell their underwriters to work faster — they gave them better tools and a better workflow. The result was faster, cheaper, and more accurate. That combination is rare, but it's achievable with the right operational focus. Luna: And they did it in a heavily regulated industry. So there's no excuse for others. Lucas: Exactly. The regulations didn't change — the process did. That's the takeaway: look at your bottlenecks, question every handoff, and ask whether the work is structured to serve the customer or your internal convenience.