Latest / Fintech Conversations with Fexingo: Banking, Payments, and Financial Software Companies / How Fintechs Use AI to Personalize Credit Card Rewards
Transcript
- Lucas: You know, Luna, I was looking at my credit card statement last night, and I realized I earned about 40,000 points last year that I'll probably never use. And it got me thinking — why are rewards still so generic? Luna: Right, it's like the airline miles problem. You earn them, then you forget, or the redemption options don't actually fit your life. Lucas: Exactly. But that's starting to change. A handful of fintechs and card networks are using AI to personalize rewards in real time. Instead of a flat one percent cash back on everything, your card might start offering you five percent at a specific coffee chain you visit every Tuesday morning. Luna: So the AI is learning your habits and serving offers on the fly. That sounds great for the consumer, but it also means the card issuer knows a lot more about you. Lucas: That's the trade-off. Let's talk about how it works under the hood. Mastercard, for instance, has a product called 'Smart Offers' that uses machine learning to analyze transaction streams and predict which merchants a cardholder is likely to engage with. Then it dynamically adjusts the rewards rate at the point of sale. Luna: So the offer is triggered before you even swipe? That's fast. Lucas: Yes. It's happening at the authorization level. Mastercard processes roughly 165 billion transactions annually — that's a massive training set. Their AI looks at location, time of day, merchant category, even the weather in some pilots. If it's raining and you usually order takeout on rainy Wednesdays, you might get a push notification offering double points at your usual Thai place. Luna: I can see why merchants love this. They're paying for a higher rewards rate, but they're getting a near-guaranteed conversion. Lucas: Exactly. It's performance-based marketing wrapped inside the payment itself. Visa has something similar called 'Visa Offers' — they've been rolling it out with several European banks. The AI there is also incorporating open banking data. So if you give permission, the system can see your full financial picture — not just what you spend on Visa cards. Luna: Wait, so it's using bank account data from other institutions to tailor the rewards? That's a big leap from just transaction history. Lucas: It is. And it's controversial. But the pitch is that by seeing your rent, your utility bills, your salary deposits, the model can infer your lifestyle much more accurately. For example, if you're consistently paying a gym membership, the AI might offer you a bonus at a health food store nearby. Luna: I guess the question is whether consumers will trust that level of data sharing. Especially after the whole data-privacy backlash of the last few years. Lucas: And that's why opt-in is critical. Both Mastercard and Visa are framing these as permission-based features. But the real test is whether the value exchange is obvious enough. If you get a 10 percent cash-back offer on something you were going to buy anyway, you might say yes to sharing more data. Luna: What about American Express? They've always been the premium rewards player. Lucas: Amex is taking a slightly different angle. They're using AI to personalize the payment experience itself. Their 'Pay It Plan It' tool — which lets you split purchases into installments or pay small charges immediately — now uses machine learning to suggest which transactions to plan and which to pay off right away. It's not exactly rewards, but it's personalization of the card's utility. Luna: So the AI is nudging you toward better financial behavior based on your spending patterns. That's more like a budgeting coach built into the card. Lucas: Right. And they're also experimenting with dynamic category bonuses. Instead of the classic 'three points on travel, two on dining,' Amex is testing offers that shift every month based on your activity. If you haven't used the card for groceries recently, you might get a temporary boost there. Luna: That seems like a smart way to increase card usage without just raising the base rate. Lucas: Exactly. It's more capital efficient for the issuer, too. They only pay higher rewards where they see a behavioral gap. Then there's the startup angle. Have you heard of Cardless? Luna: I think so — they do co-branded credit cards for specific brands, right? Lucas: Yes. But they've built an AI layer on top. Cardless partners with merchants like Atlas Obscura or a restaurant group in Miami, and the AI personalizes rewards for each cardholder based on their engagement with that brand. If you're a frequent traveler in the Atlas Obscura community, you might get bonus points for booking offbeat tours. The merchant doesn't have to guess what works — the machine learns it. Luna: So it's almost like a rewards marketplace where the issuer is acting as a matching engine between consumer preferences and merchant offers. Lucas: That's exactly the right frame. And I think that's where the whole industry is headed. The old model was: earn points, redeem them from a catalog. The new model is: the card predicts what you want and gives it to you before you ask. Luna: But doesn't that make the rewards less fun? Part of the appeal of points is the game of maximizing them. Lucas: It's a different kind of fun. Instead of hunting for the best redemption value, you get a steady stream of small, relevant wins. And for the issuer, engagement goes up because the offers feel personal. Luna: There's also a potential downside: filter bubbles. If the AI only shows you offers based on past behavior, you might never discover new categories. Your card becomes a mirror of your habits, not a window to new ones. Lucas: That's a really good point. The best systems probably need to inject some randomness — a 'wildcard' offer now and then to broaden your horizons. Visa has talked about something called 'exploration rewards' in their internal research. Luna: Let's talk about the data infrastructure required for all this. To personalize in real time, you need a lot of compute at the point of sale. Lucas: It's massive. Mastercard built a dedicated AI layer called 'Decision Intelligence' that sits between the transaction switch and the issuer. It can score an authorization request in under 30 milliseconds. That's fast enough to apply a personalized rewards rule without slowing down the checkout. Luna: And the models are getting better at handling privacy. Techniques like differential privacy and federated learning mean the raw transaction data doesn't have to leave the cardholder's device or the issuer's environment. Lucas: Right. So the model learns patterns without ever seeing the underlying personal details. That's key for regulatory compliance, especially in Europe under PSD2 and GDPR. Luna: I wonder if this trend will make traditional loyalty programs — like the ones from airlines and hotels — less relevant. If your credit card already knows you want a coffee and gives you a discount, why bother hoarding miles? Lucas: I think they'll coexist, but the balance shifts. Airlines will still offer miles for status, but the day-to-day spend might be better captured by these ai driven card perks. Some travel cards are already blending both: you earn miles, but you also get instant statement credits for things like Uber or Amazon. Luna: So the miles become the fallback, and the personalized offers become the primary value driver. Lucas: That's my bet. And it's not just about rewards. The same AI engines can power fraud detection, credit line increases, and budgeting tools. The card becomes a financial assistant, not just a payment method. Luna: Which brings us back to the trust question. Are consumers ready for that level of intimacy with their card issuer? Lucas: Early data suggests yes — if the value is clear. JPMorgan Chase reported that users of their personalized offers feature spend 18 percent more on the card within three months. That's a huge lift. Luna: Eighteen percent — that's not chump change. No wonder the networks are investing billions in this. Lucas: And what's interesting is that smaller fintechs like Cardless or Deserve are moving faster than the incumbents. They don't have legacy mainframe systems, so they can deploy new models in weeks instead of months. Luna: So the incumbents might end up buying or licensing their tech. We've seen that pattern before in fintech. Lucas: Absolutely. I wouldn't be surprised if by 2028, every major card issuer has some form of AI personalization baked in. The question is how transparent they are about it. Luna: And whether consumers can opt out without losing the core benefits. If personalized offers become the default, people who don't want to share data could end up with worse deals. Lucas: That's a real risk. Regulators are starting to look at this. The Consumer Financial Protection Bureau in the US has issued a request for information on algorithmic personalization in financial products. They're specifically asking about fairness and bias. Luna: Because if the AI learns patterns from a biased dataset, it could reinforce inequality. For example, offering better rewards to people in higher-income neighborhoods. Lucas: Exactly. Mastercard has published some research showing that their models are designed to avoid that — they explicitly include fairness constraints. But it's early days, and the auditing mechanisms are still maturing. Luna: So where does this leave the average consumer? What's the one thing they should know about their credit card rewards right now? Lucas: I'd say: check whether your card issuer offers personalized deals. Many do, but they're often opt-in. If you're willing to share some data, you could be leaving money on the table. Especially if you're a consistent spender in a few categories. Luna: And if these conversations are useful for what you're building or running, we'd love your support. It's what keeps Fintech Conversations ad-free and focused on the substance. You can find us at buy me a coffee dot com slash fexingo. Lucas: We appreciate that. And it really does help us keep digging into topics like this — the ones that are shaping how we interact with money every day. So, thanks for considering. Luna: Alright, let's wrap up with one more thought. I'm curious — do you think AI personalization will eventually eliminate the need for rewards programs altogether? Lucas: That's the billion-dollar question. I think rewards programs will evolve into something more like a dynamic pricing engine for consumer finance. The points will still exist, but they'll be just one part of a much richer, more responsive system. And the winners will be the issuers who make that system feel effortless and trustworthy. Luna: Effortless and trustworthy. That's a good bar to aim for. Thanks for the deep dive, Lucas. Lucas: Thanks, Luna. And thanks to everyone listening. See you next time.