Latest / The Operator's Manual with Fexingo: Daily Decisions, Tactical Wins, and Business Execution / How One Bank Cut Loan Approval Time 80 Percent With a Single Rule Change
Transcript
- Lucas: So there's this regional bank — I'm not naming them because they're private and they don't want the attention — but they made one change to their small-business loan process last year that cut approval time from eighteen days to under seventy-two hours. Luna: Eighteen days to three days? That's a massive jump. What was the change? Lucas: They eliminated the manual underwriting review for any loan under two hundred fifty thousand dollars. That's it. One rule. They replaced it with an automated decision engine that checks five criteria — credit score, cash flow, time in business, industry risk, and the owner's personal guarantee. Luna: So they took the human underwriter out of the loop for the smaller loans. Was that a hard sell internally? Lucas: It was brutal. The head of credit risk almost quit. He had been at the bank for twenty-two years, and his whole identity was built around saying 'no' to risky loans. The data was on the side of the change, but culture — that was the real fight. Luna: What data did they use to make the case? Lucas: They went back and looked at every small-business loan they'd approved manually over the previous five years. Loans under two fifty. They found that the manual underwriter — the human — changed the outcome in only three percent of cases. Ninety-seven percent of the time, the automated rules would have made the same decision. Luna: So basically the human was mostly just stamping paperwork, not adding value. Lucas: Exactly. And worse — the three percent where they did change the decision? They were actually more likely to approve a loan that later defaulted. The model was more conservative than the humans on the margin. Luna: So the manual step wasn't just slow — it was making the portfolio riskier. Lucas: That's the part that surprised everyone. The head of credit ran the numbers himself and came back two weeks later saying — I'm paraphrasing — 'I think we've been making things worse.' Once he saw it, he became the biggest champion of the change. Luna: What were the actual results after they flipped the switch? Lucas: First quarter with the new rule: median approval time dropped from eighteen days to fifty-four hours. Loan volume went up forty percent — not because they approved riskier loans, but because the process didn't lose applications. Previously, about one in five applicants would just give up during the waiting period. Luna: So the eighteen-day process was actively driving away customers who would have qualified. Lucas: Right. And here's the kicker — the default rate on those automated approvals after twelve months was actually lower than the historical average on manually underwritten loans. The head of credit presented that at a board meeting and said, 'We were the problem.' Luna: That is a powerful sentence for a career risk officer to say out loud. Lucas: It changed the conversation. Suddenly the conversation wasn't about 'are we loosening standards?' It was about 'why are we still manually reviewing anything under two fifty?' Luna: And did they expand the rule? Lucas: They did. Second phase, they raised the threshold to five hundred thousand. Same data-driven approach — they retroactively tested the model on larger loans and found it performed well. But they kept the manual step for anything above five hundred thousand, because those loans have more complex structures — real estate collateral, partnership guarantees, that kind of thing. Luna: So the principle was: automate where the decision is straightforward, keep the human where complexity is high. Lucas: Exactly. And the beautiful thing is, the underwriters didn't lose their jobs. They got redeployed to the complex cases — the ones where their judgment actually matters. The bank went from having five underwriters each handling a hundred and fifty applications a month to each handling maybe thirty, but those thirty were bigger, more interesting, and the bank made more money on them. Luna: So the work got more valuable for the employees too. Lucas: Yes. And that's the part I think a lot of companies miss when they think about automation. They think 'machine replaces human.' But in this case, the machine replaced a low-value task, and the human got upgraded to high-value work. Luna: What about the customer experience? Did they measure that? Lucas: They did. Net promoter score for the small-business loan process went from minus twelve — meaning more detractors than promoters — to plus thirty-four within six months. The main complaint had been 'it takes too long and I never hear back.' The automated system sends an approval or decline within seconds. And if it's a decline, it tells you why — 'your debt to income ratio is above our threshold' — so the applicant can actually do something about it. Luna: Which is exactly the kind of friction reduction we talk about on this show. And it makes me think — if these conversations have moved your work forward in some small way, listener support is what keeps this show ad-free and independent. Lucas: Yeah, it's a simple ask. If you've gotten value from episodes like this one — the kind where a single rule change saves weeks of time — you can support us at buy me a coffee dot com slash fexingo. No pressure, no perks, just knowing you're keeping the lights on. Luna: And we mean that. Now, back to the bank — the other interesting outcome was how it changed their risk culture. Lucas: Right. The head of credit started publishing a monthly 'rules versus judgment' report. He'd pick a random sample of approved loans and compare what the automated engine would have decided versus what the underwriter decided. It became a tool for continuous learning — spotting cases where the model was missing something and cases where the human was biased. Luna: So the rule change didn't just improve speed — it created a feedback loop for the whole underwriting team. Lucas: Exactly. And that feedback loop is probably worth more than the speed gain. Because now every month they're testing their own assumptions against data. The CEO told me the culture shifted from 'we've always done it this way' to 'prove it with data.' Luna: That's the kind of cultural change that outlasts any process change. Lucas: One last number that stuck with me: the bank estimated that the manual review step cost them about eight hundred thousand dollars a year in underwriter salary and overhead for the small-loan process. After the change, that cost dropped to about a hundred thousand — mostly system maintenance. So they saved seven hundred grand a year, plus the revenue from the forty percent more loans. Luna: Seven hundred thousand in direct cost, plus incremental revenue. On a single rule change. Lucas: Yeah. And the rule change itself was basically deleting one line in a policy manual. No software purchase, no consultants, no new headcount. Just a decision to stop doing something that wasn't adding value. Luna: That's the kind of episode that makes you look at your own process and ask: what are we doing manually that we could test with data? Lucas: That's the question. And if you find something, the hard part isn't the data — it's the conversation with the person who has been doing it for twenty-two years.