Will LLMs Solve Distribution For The Cash Flow Lenders?
Updating my thoughts on an October 2024 Free Toaster Editorial
A year and a half ago, I wrote a piece called Where’s The Consumer Demand For Cash Flow Underwriting (CFU)?
The thesis was simple. The B2B CFU providers were buzzy and raising big money. But I couldn’t find a single cash-flow-based lender who was actually killing it. And the reason came down to one word: distribution.
Selling a CFU loan product was high friction. The second a consumer hit “please give us your bank login,” they bailed. That kills your application rate. A lower application rate kills your throughput. And lower throughput means the affiliate marketplaces favor the traditional, credit-bureau-only lenders every time they have a choice.
I still think all of that was true then. And honestly, most of it is still true today.
But I’m starting to change my mind about where this goes long-term. So let me tell you what shifted.
First, two things I should have hit harder
When I wrote the original piece, I framed CFU mostly as a primary underwriting decision. New customer comes in, you run them on cash flow, you approve or decline.
Talking to folks much closer to the chief risk officer seat than I am, I’ve since learned the two use cases that are actually working today aren’t that at all. And one of them is extremely relevant to affiliate marketing.
1. Second look
Customer comes through your funnel. You decision them on traditional credit bureau data. You decline.
Instead of letting that lead die, you say: “We can’t approve you on your bureau alone. But if you can show us your cash flow looks healthy, we might be able to.”
Some of those declines will go through Plaid. And some of those will then get approved.
Here’s why this matters for anyone running an affiliate program. You just did three things at once:
You monetized leads that were dead weight before.
You drove up your net approval rate, which lets you pay higher bounties to your publisher partners on the same lead pool.
You improved your funnel enough that you can climb the rankings, win better placement, and pull more volume and higher quality customers.
That’s the flywheel. And second look spins it without forcing every consumer through the friction upfront. You only ask for the bank login from the people you’ve already told “no” the traditional way.
2. Customer management
This one is less relevant to affiliates, but worth naming. Once you’ve already booked the customer, friction stops being mission critical. You’ve got them.
So now you can use cash flow data for things like credit line increases and repricing in the consumer’s favor. Picture someone in their online banking app checking their balance, asking for a CLI. In that moment you say, “To get you the best terms, I’d love to see what’s going on in your account.” They link Plaid. You see something positive. You give them the increase.
Neither of these was floating top of mind when I wrote the original piece, but they’re great practical examples of how lenders are using cash flow data today.
The three things I said had to change
At the end of that article, I said we didn’t have a good answer for how to get CFU into consumers’ hands at scale, but that the answer would include three things:
Distribution channels that can use cash flow data at the marketing stage to identify who qualifies
Reducing the application friction in consumer-permissioned access to checking accounts
Educating consumers on the value and safety of sharing their cash flow data
Let me go one by one, because all three have moved.
1. Distribution: enter the LLM
This is the big one, and it’s the thing that prompted me to write today’s piece.
Back then, the only channel I could point to that used cash flow data at the marketing stage was MoneyLion. That was basically it.
Now look at what’s happening with partnerships like ChatGPT and Plaid. You can connect your Plaid to the LLM and start asking it anything. How much did I spend on groceries last month? How does that compare to last year? How much am I saving? Where is my money actually going?
This is everything Mint.com tried to do, except it’s not a clunky dashboard. It’s personalized, contextual, and conversational. And I think it all moves to voice. I’ll fire up Claude and ask, “How much did I spend on gas two years ago, and what do I think I’ll spend this year given my driving has changed?” and get a sharp answer in seconds.
That reframes the entire distribution question. It used to be: can your bank convince you to link Plaid? Can Credit Karma? Now the LLMs have a genuinely compelling reason for the consumer to link their accounts, because the more context you give the model, the better it serves you.
This is what Nick and I have been calling the personal finance super app. The data plumbing for cash flow is getting laid down by consumers themselves, voluntarily, because they want better answers about their own money.
2. Friction: Plaid is starting to remember you
The second thing has moved too. Plaid is rolling out what’s effectively a returning-user shortcut. If you’ve linked through Plaid before, you put in your phone number, they recognize you, they text you a code to confirm it’s you, and you’re in. No hunting for usernames and passwords across a dozen banks.
I’m hearing directly from lenders that conversion rates are up as a result. That’s the whole ballgame for a funnel that used to scare customers at the login step.
3. Education: this isn’t an education play anymore
I used to think we’d need to teach consumers why sharing cash flow data was valuable and safe. I don’t think that’s the work anymore.
The LLMs are making it abundantly obvious on their own. The more you share, the more context you give, the better the answers you get back. Consumers are learning the value proposition by using the product, not by reading a brochure about it.
The open question
So where does this land?
The thing I’m genuinely unsure about is who owns the consumer relationship. Does ChatGPT own it? Does the bank build its own LLM layer that does this natively? Does a marketplace with an existing brand in consumer finance lean in? Or does someone go AI-native, build the whole thing from scratch, and earn the consumer’s trust as a newcomer?
Jury’s out. Could be OpenAI, Anthropic, Perplexity, or a finance-native brand we haven’t met yet. I don’t know who plays that role.
But I’m now convinced the role gets played.
So what?
If you read what I wrote a year and a half ago and concluded “as a cash-flow-based underwriter, I’m going to struggle to get distribution from the affiliate marketplaces,” you were right. You’re still right today.
What I no longer believe is that it stays that way.
If you’re a lender who isn’t all-in on cash flow underwriting, now is the time to start watching the board:
Where is this data starting to live?
Which surfaces have it?
How many consumers have plugged in?
And how good is the conversion path from “I’m asking my LLM about my groceries” to “I’m taking out a new personal loan”?
That last one is the real open question. It’s possible consumers just do budgeting and personal finance chats that never turn into a new product. But I’d bet a lot of them do convert, because the natural next move after understanding your money is optimizing it. Better rewards. Lower rates on your loans. Consolidating debt. Those are exactly the products a lender brings to the table.
I won’t put a tight timeline on it. But I’d bet that in about three years, this is a materially different market.
I’m going to be watching it closely, and I’ll report out here as it develops.
If you’re working on getting a cash flow product working in affiliate channels, I’d love to compare notes.
Please support our Newsletter by sharing this edition with a colleague!
And, consider checking out our sponsors:
Need an affiliate marketing platform/network? Chat with Fintel Connect
Need to improve application page conversions? Chat with Spinwheel
Need help with performance marketing? Chat with Bulldog Media Group
Need to grow in big marketplaces like Credit Karma? Chat with New Market Growth
Need to hire top-tier talent in fintech or lending? Chat with CTB1