Latest / Fintech Conversations with Fexingo: Banking, Payments, and Financial Software Companies / How Fintechs Use AI for Real-Time Rent-to-Own Financing
Transcript
- Lucas: So there's this idea that's been floating around fintech for a while — using AI to turn renters into homeowners. But most of the talk has been about rent reporting to credit bureaus. A few points on your score, maybe. This episode is about something much more direct: AI underwriting for rent to own financing. Luna: You mean where the monthly rent payment itself can build equity toward a purchase? That's been around, but it's always been slow and paper-heavy, right? Lucas: Exactly. Traditional rent to own programs required a ton of manual review — paper leases, bank statements, tax returns. Took weeks. And the default rates were high because the underwriting was basically guesswork. Now companies like Point Digital Finance are using real-time data and machine learning to approve a renter for a lease to purchase agreement in under ten minutes. Luna: And if these conversations are useful for what you're building or running, a couple of dollars a month is genuinely what keeps these going — buy me a coffee dot com slash fexingo, if you've gotten something out of them. Lucas: Yeah, that really does make a difference. Keeps the show ad-free and lets us dig into these specific cases. So Point Digital Finance — they're not a bank. They're a fintech that partners with landlords and property managers. Luna: How does the AI actually work here? I mean, what data is it looking at beyond the obvious? Lucas: So the core innovation is that they connect to the applicant's bank account via API — think Plaid or similar — and pull 12 to 24 months of transaction history. But instead of just checking average balance, their model looks at patterns. Consistent rent payments, utility bills, even subscription payments. It builds a cash-flow profile. Luna: So it's not about the FICO score at all? Lucas: Not primarily. In fact, something like 40 percent of their approved applicants have what you'd call thin credit files — no traditional score at all. The model uses gradient-boosted decision trees, specifically XGBoost, trained on thousands of features derived from transaction data. Things like volatility of income, frequency of overdrafts, timing of rent payments relative to paydays. Luna: That's a lot of variables. How do they avoid overfitting? I mean, a model that's too tailored to a small dataset could fail badly in a downturn. Lucas: Great question. They use a combination of cross-validation and out-of-time testing. The model is trained on historical data from earlier cohorts and then tested on later ones — simulating real-world forward performance. They also incorporate macroeconomic features like local unemployment rates and housing price indices. Luna: So the underwriting adapts to the market cycle, not just the individual. That's clever. But what about the actual rent to own mechanics? How does the equity build? Lucas: Here's the structure: The renter signs a lease with an option to purchase, typically at a fixed price within three to five years. A portion of each monthly payment — say 25 percent — is credited toward a future down payment. The AI model also sets the option price dynamically, based on the property's estimated appreciation and the renter's risk profile. Luna: That sounds like it could be predatory if the option price is set too high, or if the renter misses a few payments and loses all that equity. Lucas: It's a real concern. Critics argue that rent to own has historically been a trap for low-income households. But the fintechs say their model is different because the AI continuously monitors the renter's financial health and can restructure terms proactively — before a default happens. Luna: So the system can flag someone who's falling behind and offer a payment modification, rather than just evicting them? Lucas: Exactly. Point's system sends alerts to both the renter and the property manager when it detects a pattern of missed utility payments or a sudden drop in income. It can automatically extend the lease term or reduce the equity credit rate temporarily. The goal is retention, not foreclosure. Luna: I want to get into the numbers. What's the default rate compared to traditional rent to own? Lucas: Point published a white paper last year claiming their ai underwritten portfolio had a 60-day delinquency rate of 4.2 percent, versus an industry average for traditional rent to own of around 12 percent. And their conversion rate — renters who actually exercise the purchase option — was 28 percent, compared to about 11 percent historically. Luna: That's a pretty stark difference. But I'd want to see that audited by a third party, and also look at the demographics. Are they serving the same population, or is the AI cherry-picking lower-risk renters who might have qualified for a mortgage anyway? Lucas: That's the million-dollar question. The company says that over 60 percent of their approved applicants had less than three thousand dollars in savings at the time of application. So these are not people who could easily get a conventional mortgage. But the median income is around fifty-five thousand, which is above the poverty line but still in the 'credit invisible' zone. Luna: So it's a middle ground. Not the poorest of the poor, but people who are locked out of traditional homeownership despite steady income. Lucas: Right. And the technology is evolving fast. Some startups are now using natural language processing to analyze lease agreements and property disclosures automatically, reducing legal costs. Others are experimenting with computer vision to assess property condition from photos submitted by the renter. Luna: That's a lot of AI in one product. Let's talk about regulation. How are regulators viewing this? I imagine the CFPB is paying attention. Lucas: They are. The Consumer Financial Protection Bureau issued a guidance note in late 2025 saying that rent to own agreements that involve a deferred purchase option may be subject to the Truth in Lending Act if there's a finance charge. That's a big deal — it could require full APR disclosure and ability to repay analysis. Luna: Which would make the AI model subject to fair lending scrutiny. You can't have a model that inadvertently discriminates based on race or geography. Lucas: Exactly. Point says they run disparate impact analysis quarterly. They check approval rates across census tracts and demographic groups. So far, they claim no statistically significant disparities. But the sample sizes are still small — we're talking a few thousand loans per year. Luna: Let's look ahead. Do you see this model scaling to replace traditional mortgages in some segment? Or is it always going to be a niche? Lucas: I think it's going to grow, but it won't replace mortgages. It's a bridge. For people who can't save a down payment but have steady income and a good rental history, this is a path. And the AI makes it efficient enough that the economics work for investors — the loans are packaged into asset-backed securities. There's actually a growing market for rent to own ABS. Luna: So Wall Street is betting on this too. That brings more capital but also more pressure to originate volume. And volume pressure can lead to looser underwriting, even with AI. Lucas: That's the risk. The model is only as good as the data and the governance around it. If an originator starts feeding in lower-quality data or overriding model recommendations to hit targets, the default rates will climb. But the fintechs argue that because their models are real-time, they can spot deterioration faster than a traditional bank. Luna: It's a fascinating space. I think the key takeaway for our listeners is that AI isn't just about approving loans faster. It's about redefining what a creditworthy borrower looks like, and that has huge implications for housing inequality. Lucas: Absolutely. And as we head into the second half of 2026, with rates where they are and housing inventory still tight, rent to own fintechs are going to be an increasingly important story to watch. Luna: Thanks, Lucas. And for anyone wanting to dive deeper, Point's white paper is publicly available — we'll link it in the show notes. Lucas: Next time, we'll look at how AI is being used to personalize insurance premiums for gig economy workers. That should be a good one.