Latest / Fintech Conversations with Fexingo: Banking, Payments, and Financial Software Companies / How Fintechs Use AI for Real-Time Supply Chain Financing
Transcript
- Lucas: So supply chain financing — it's one of those areas where fintech has been quietly revolutionary, but most people outside of treasury departments have no idea how much it's changed over the past two years. Luna: Yeah, when I hear 'supply chain finance' I still think of old-school invoice factoring — a supplier sells its receivables at a discount and waits for payment. What's the AI twist? Lucas: Right, that's exactly the old model. Traditional factoring is reactive — a supplier has a cash crunch, they go to a bank or factor, and they get maybe 70 to 80 percent of the invoice value upfront, but it's slow and expensive. What we're seeing now is ai driven dynamic discounting and real-time risk scoring. Lucas: Think of a platform like Taulia or C2FO. They connect to a buyer's ERP system — SAP, Oracle, whatever — and they can see every approved invoice in real time. Then their AI scores each supplier's risk profile and payment history, and offers the buyer an optimal discount rate to pay early. Luna: So the buyer gets a discount for paying early, and the supplier gets cash faster. But the AI is deciding what that discount should be, dynamically? Lucas: Exactly. It's not a fixed 2 percent net 30 anymore. The AI might say: for this supplier with a strong track record, offer 1.5 percent if paid in 10 days. For a newer supplier with higher risk, maybe 3 percent. And it adjusts based on real-time data — if the supplier just got a credit downgrade, the discount might widen that afternoon. Luna: That's pretty wild. How real-time are we talking? Like, same-day settlement? Lucas: In some cases, yes. There are networks now where a supplier can request early payment at, say, 10 AM, and the funds hit their account by 2 PM. The AI has already calculated the buyer's available working capital, the supplier's risk score, and the optimal discount — all automated. Lucas: Take a specific example I came across recently. A mid-sized manufacturer in the Midwest with about 500 suppliers. Before they implemented this, their average days payable outstanding — how long they took to pay invoices — was 45 days. After integrating with an ai driven platform, they brought it down to 12 days for suppliers who opted into dynamic discounting. Luna: Twelve days? That's a huge shift. What was the incentive for the buyer? They're paying earlier, so they're out cash sooner. Lucas: The incentive is the discount. They earned an average of 2.8 percent annualized return on the early payments, because the discounts they captured were higher than what they could get on short-term cash investments. And for the suppliers, it meant they didn't have to go to a factor and lose 10 or 15 percent. Luna: So it's a win-win. But this only works if the AI is accurate on risk, right? If it misprices a discount, either the buyer leaves money on the table or the supplier gets a raw deal. Lucas: That's the key. The risk models use not just traditional credit data but also transaction history from the platform itself, plus external signals like macroeconomic indicators, industry trends, even news sentiment. C2FO, for instance, claims their AI can predict supplier default risk with over 90 percent accuracy within a 30-day window. Luna: Ninety percent — that's impressive if true. But I wonder about the data privacy side. The platform is sitting inside the buyer's ERP, seeing every invoice, every payment term. That's sensitive. Lucas: It is. And that's actually one of the barriers to adoption. Some large buyers are reluctant to give a third-party fintech that level of access. But the platforms argue that they only see invoice-level data, not bank accounts or customer lists, and they sign strict NDAs. Plus, the value proposition is strong enough that many are overcoming that hesitation. Lucas: And it's not just standalone fintechs. The big banks are piling in too. JPMorgan launched a real-time supply chain finance product last year that uses AI to dynamically adjust pricing. Citibank has something similar through their trade finance unit. They're trying to defend their corporate banking relationships against the upstarts. Luna: So who's winning? The fintechs or the banks? Lucas: Right now, it's a fragmented market. Taulia and C2FO each handle tens of billions in transaction volume annually. But JPMorgan has the advantage of existing relationships with the Fortune 500. The fintechs have speed and better UX. I think we'll see more partnerships — banks providing the balance sheet, fintechs providing the AI layer. Luna: Yeah, that makes sense. And this is all happening because of higher interest rates, right? When money is cheap, companies don't care as much about optimizing payment timing. But at 5 percent rates, every day of float matters. Lucas: Absolutely. Since rates started climbing in 2022, the working capital optimization space has exploded. Companies that used to pay invoices on net 60 terms are now actively looking to accelerate payments to suppliers in exchange for discounts, because the return on that cash is better than leaving it in a money market fund. Lucas: And the AI piece is crucial because you can't manually optimize thousands of invoices across hundreds of suppliers in real time. It has to be algorithmic. Luna: So what's the next frontier? Where does this go from here? Lucas: I think we'll see AI start to predict not just optimal discounts but also suggest changes to payment terms dynamically. So a supplier with a perfect payment history might get offered net 15 instead of net 30, based on the AI's confidence. And on the buyer side, the AI could recommend which invoices to pay early based on the supplier's cash need and the buyer's own cash position. Lucas: Some platforms are already experimenting with using machine learning to forecast a supplier's cash flow needs before the supplier even requests early payment. That's proactive rather than reactive. Luna: That's fascinating — almost like a financial health monitor for the supply chain. And it makes me think: this kind of ai driven financing could really help smaller suppliers who don't have access to traditional bank credit. Lucas: Exactly. That's the inclusive angle. A small supplier with no credit history but a solid payment record on the platform can get better terms than they ever could from a bank. The AI sees their actual transaction behavior, not just a credit score. Luna: And this is only possible because of the real-time data integration. It's a good example of fintech solving a real operational problem. Lucas: Before we go deeper — and I know this is a bit meta — but conversations like this, where we get to explore how technology is reshaping business finance, they take a lot of research and production. And honestly, the reason we can keep doing this ad-free is because a small group of listeners chips in monthly at buy me a coffee dot com slash fexingo. Luna: Yeah, it's true. We don't run ads, we don't have sponsors pushing a narrative. It's just us digging into topics we find genuinely interesting. And that model works because of listener support. Lucas: So if you find value in these episodes, consider being part of that. It literally funds the research time and the tools we use. No pressure, just something to think about. Luna: Alright, back to supply chains. Lucas, you mentioned proactive cash flow forecasting. How far off is that from being mainstream? Lucas: I'd say it's already happening in pilot programs with large buyers. A few automotive manufacturers and retailers are testing it. The AI looks at historical payment patterns, seasonality, and even weather data — because a hurricane in the Gulf can affect raw material deliveries, and the system adjusts financing offers accordingly. Luna: Weather data? That's a level of granularity I hadn't considered. Lucas: It sounds futuristic, but it's really just another data feed. The more signals the AI can incorporate, the better it gets at predicting when a supplier will need cash and what discount to offer. And all of this happens in milliseconds. Luna: So what's the biggest obstacle to wider adoption? Is it the ERP integration? The trust issue? Lucas: I think it's a combination. ERP integration is getting easier — most modern ERPs have APIs. But the cultural shift is harder. Treasury teams are used to predictable payment cycles. Letting an AI decide when to pay and at what discount requires a leap of faith. Plus, there's regulatory uncertainty — if the AI makes a mistake and a supplier goes under because they relied on an early payment that didn't come, who's liable? Luna: Right, the liability question is huge. And it's not just about the discount — it's about the supplier's entire cash flow planning. Lucas: Exactly. So we'll likely see a hybrid model for a while: AI recommends, humans approve. But as the track record builds, I expect more automation. The ROI is just too compelling to ignore. Luna: Well, this has been a great deep dive. I feel like I finally understand what 'dynamic discounting' actually means. Lucas: Glad to hear it. It's one of those fintech use cases that's quietly transforming B2B finance, and I think we'll see a lot more innovation here in the next year.