Karoo Is Quietly Building a Better Loan Product for the People Banks Ignore

The Macro: Subprime Lending Is a Trillion-Dollar Mess

About 100 million Americans have credit scores below 670. Banks don’t want to lend to them. The risk models don’t work, the margins are thin, and the regulatory scrutiny is intense. So these borrowers get pushed toward a shadow market of high-interest personal loans, payday lenders, and buy-now-pay-later products that are payday loans wearing a blazer.

The numbers are staggering. The subprime personal loan market in the U.S. exceeds $200 billion. Average APRs for near-prime borrowers (620-659 credit scores) run 20-36%. For deep subprime, rates can exceed 100% APR when you factor in fees. That’s not lending. That’s extraction.

The incumbents in this space are not incentivized to fix it. Companies like Avant, OppFi, and Elevate Credit have built profitable businesses around high-rate lending to underserved borrowers. Their risk models price in high default rates and compensate with high interest rates. The borrowers who don’t default end up subsidizing the ones who do, at rates that would make a credit card blush.

The fintech wave was supposed to fix this. SoFi, LendingClub, Prosper. They all started with promises about democratizing credit. Most of them drifted upmarket toward prime borrowers where the unit economics are simpler. The underserved market remained underserved.

The AI angle is interesting here because traditional credit scoring is notoriously bad at evaluating borrowers without conventional credit histories. FICO scores miss gig workers, immigrants, young adults, and anyone who has been financially responsible but invisible to the credit bureaus. If you can build a better underwriting model using alternative data and machine learning, you can price risk more accurately, charge lower rates, and still maintain a profitable loan book.

The Micro: Lendable Alumni Who Know the Lending Stack Cold

Karoo is building an AI-native consumer lending platform targeting the space between prime lending and payday loans. Their pitch is simple: the market is inefficient, incumbents are price gouging, and better technology means better pricing for borrowers.

The founding team comes from Lendable, which is itself an interesting signal. Lendable is one of the larger fintech lending platforms globally. Jeanot Dawson is the CEO, previously Head of U.S. Loans at Lendable. Hugo Markland is the Chief Capital Officer, an early employee at Lendable who led UK capital markets and helped raise billions in funding. Oliver Lambson is the CTO, whose bio claims they went from zero lines of code to first loan issued in 45 days.

That 45-day stat is worth paying attention to. In consumer lending, the distance from “idea” to “actual loan disbursed” is usually measured in quarters, not days. Lending requires underwriting models, bank partnerships, regulatory compliance, identity verification, payment processing, and fraud prevention. Shipping all of that in 45 days suggests either extreme velocity or an extremely minimal MVP. Either way, they’re moving fast.

They’re part of YC’s Winter 2025 batch, based in Toronto. The website shows loans from $500 to $15,000 with an emphasis on speed, transparency, and no hidden fees. They’re currently live in Canada with a separate U.S. operation at karoo.loans, which is the market where the real opportunity sits.

The “AI native” framing means the underwriting model is built on machine learning from the ground up rather than retrofitted onto traditional credit scoring. If that model can identify good borrowers that FICO misses, Karoo can offer lower rates and still maintain healthy margins. That’s the core bet.

The Verdict

I think Karoo has the right team for this specific problem. Consumer lending is one of those spaces where domain expertise matters enormously. Regulatory knowledge, capital markets relationships, and experience managing loan portfolios through credit cycles are not things you can fake. The Lendable pedigree gives them all three.

The risk is capital. Consumer lending is a capital-intensive business. You need money to lend before you can earn from lending. Securing debt facilities at favorable rates as a startup is hard, and the terms you get on your capital directly determine whether you can price competitively. Markland’s capital markets background is the hedge here. If he can source cheap capital the way he did at Lendable, the unit economics work. If they’re stuck with expensive warehouse lines, the rate advantage over incumbents evaporates.

The regulatory risk is also real. Lending to underserved borrowers attracts attention from state regulators and the CFPB. Any lending platform in this space needs bulletproof compliance infrastructure from day one, not as an afterthought.

In 30 days, I want to see loan volume and default rates. Early cohort data is the most important signal in lending. In 60 days, the question is whether the AI underwriting model is actually outperforming FICO for this borrower segment. If it’s not, they’re just another subprime lender with a nicer website. In 90 days, I’d want to see the U.S. launch trajectory. Canada is a good testing ground, but the addressable market in the U.S. is 10x larger. The team has the background to build this. The question is whether “better underwriting” is enough of an edge when the incumbents have billions in deployed capital and years of default data.