{ad_logo_primary}

Quick favor: Our records indicate that you aren’t opening this email. But records can be wrong. Please click here if you’d like to remain subscribed to Fintech Takes. 

Happy Wednesday, Fintech Listeners!

It’s been a great week so far. 

Among many other things, my wife and I finished watching the very short first season of The Madison. As a native Montanan who both enjoys and despises Taylor Sheridan’s obsession with my home state, I feel it’s my duty to give you my quick thoughts (mild spoilers):

• The show wasn’t at all what I expected. It’s basically a six-episode meditation on grief, which is just so different from anything Sheridan has done up to this point.
• Michelle Pfeiffer is, low-key, the villain of the show? That was my wife’s conclusion by the end of the season, and I think I agree.
• I have owned all of the Patrick J. Adams stock since season one of Suits, and I continue to be happy with that investment.
• There’s a certain nobility to the art and ritual of fly fishing, but it’s also a silly and deeply difficult activity. There’s no way Michelle Pfeiffer’s character could master it that quickly.
• Taylor Sheridan’s seething contempt for New York and California is occasionally amusing to me, but I’m guessing it makes his shows a tough watch for the residents of those states.  

But enough TV criticism. Let’s talk about the latest Fintech Takes podcast episode!

— Alex

Was this email forwarded to you?

3 BIG IDEAS FROM THE PODCAST

This week is a real treat. I was delighted to be joined by Kiran Aware (Chief Consumer Credit Officer at LendingClub) and Michelle Young (Credit Product Lead at Plaid) for our first Facing Credit episode of 2026.

Quick housekeeping: if you haven't caught an episode of Facing Credit yet, it's our newest show within the Fintech Takes podcast feed, and it’s where we unpack what's actually happening in lending right now (if you enjoy a sense of continuity, check out our last episode here, our last last episode here, and very first episode here).

In today’s episode, we cover the macro environment, the death of "alternative data" as a meaningful concept, organizational trade-offs (you’ll be at the edge of your seat, I swear), and where AI fits into all of this in both the near and long term.

And read below for my three big ideas... 

Most people talk about the credit risk underwriting process by breaking it down into its constituent pieces and parts: analytic models, data sources, decision trees, manual review queues, and the like.   

This is a logical (if incremental) way to think about it, but it’s not the only way to think about it. And at a time of accelerating change in our industry, it’s arguably not the best way to think about it.

In our discussion, Kiran offered a different perspective. Credit underwriting isn’t just the sum of a bunch of individual components. It’s a system. An operating system.

This framing requires more holistic thinking. You can’t focus on optimizing a single component, like an analytic model. You also have to think about the data feeding that model and the policies, pricing logic, verification workflows, fraud controls, and servicing and collections capabilities stacked on top of it. 

As Kiran said in our conversation, this way of thinking is much more difficult, but it’s also far more valuable. It forces you to stop optimizing the performance of individual components and to start asking, “What should the next version of this system look like?”

At LendingClub, the operating system is built on nearly two decades of credit decisions across 100 million applications. From that foundation, they built what Kiran described as a hyper-customized machine learning ecosystem that’s wired into multiple models simultaneously, each one optimizing a different slice of the consumer credit lifecycle.

The point Kiran makes is that portfolio-level outcomes matter more than improvements to individual components. A lender can improve one model and still produce a worse business if the conversion rate drops or fraudsters find a new opening. Once you think at a system level, the unit of analysis changes. You stop asking whether one model improved, and start asking whether the outcome of the entire portfolio improved.

However, the system Kiran describes only works if the data feeding it is comprehensive. For lenders across every segment (mortgage, credit card, auto, etc.), that’s a concern that Michelle keeps hearing. 

And that brings us to …

Take two borrowers that look identical on paper, more or less (same FICO score, same income, same outstanding debt). A traditional credit underwriting system would treat them identically.

However, the paths they took to get here might tell two different stories.

Say one of those borrowers arrived at today’s snapshot after years of consistently responsible financial behavior, following an unanticipated financial crisis that knocked them off course (job loss and unexpected medical expenses are two common examples). And say the other borrower has been slowly accumulating more debt and increasingly making small mistakes over the same time frame. You have the same profile on paper, but opposite trajectories. As Kiran put it, you need to tune your credit system to understand where the consumer is going, not where they stand.

A system built on static snapshots will approve both borrowers at similar terms, even though one of them is a meaningfully worse credit risk.

You know where I’m going with this. 

Consumer-permissioned cash flow data can help with this problem! Collecting self-stated income at origination is useful. Knowing what a consumer’s income was six months ago, what it is today, and its characteristics (variety, stability, etc.) over that time is much more useful.

I obviously had to ask Kiran and Michelle about AI, and where they see it being used in the short-term and where it might add value or create new challenges over the long run.

LendingClub processes 90% of loans without human intervention. For the 10% that require it, Kiran believes that AI embedded in the fraud and verification workflow can give consumers real-time feedback on whether what they're uploading meets requirements (a process that could take days can now take minutes).

At Plaid, the value shows up in transaction categorization at scale: deploying LLMs where volume and complexity create ambiguous patterns that a rules-based system would misread.

Kiran's longer-horizon view is that credit systems will eventually generate new signals autonomously. Today, a credit analyst identifies a pattern, builds attributes and models to leverage its predictive power, and implements them in production. His future version has AI flagging behavioral shifts in real time and feeding them back into the system without waiting for a human to notice. In this way, the system would become more self-optimizing over time.

Michelle's closing point was one worth sitting with. Most of the conversation around AI in lending focuses on what lenders can do with the technology. She astutely pointed out the importance of AI’s impact on the other side of the transaction.  

Fraudsters are already using it to attack lending workflows in ways that legacy fraud prevention tools are ill-equipped to stop. But the more interesting question is what happens when legitimate customers start using AI?

Imagine a borrower using an AI agent to make financial decisions on her behalf. Imagine she experiences a disruption in her cash flow that requires her to strategically default. A human, in that instance, may prioritize certain obligations based on brand affinity or the length of time she’s had a relationship with a specific lender. An AI agent, working on the human’s behalf, may take a very different approach to that prioritization decision. 

This isn’t science fiction. It's a logical extension of where agentic AI is heading, and one that will challenge many of the foundational assumptions on which our credit system is built.

WHAT I'M LISTENING TO

Kathy Kraninger, the current president and CEO of the Florida Bankers Association and the former Director of the CFPB, obviously has a highly relevant point of view on topics like consumer protection, fraud, and AI. 

So, I’m really glad Rob had her on the podcast!

I love it when finance and the NBA collide! This is a fun episode about how to fix tanking in the NBA.

Credit is trust. And trust, as StellarFi founder and CEO Lamine Zarrad says, is contextual. In Episode 3, we explore why underwriting should work more like an iterative game than a one-time approve or decline decision, where more consumers can start somewhere, prove their way forward, and be understood via a fuller picture of their financial lives.

*This rec is brought to you by one of our fantastic brand partners.

Thanks for the read! Let me know what you thought by replying back to this email. 

— Alex  

Join 2,444 other finance and fintech leaders in the Fintech Takes Network

The conversation doesn't have to stop here
Keep learning and connecting in the Fintech Takes Network

EVENTS  |  FEED  |  LIBRARY  |  DIRECTORY

Get your brand in front of 60,900+ fintech and banking executives.

Workweek Media Inc. 
1023 Springdale Road, STE 9E
Austin, TX 78721

Want to ruin my day? Unsubscribe.