AI Shifts Underwriting From the Person to the Purchase

 | 

And as latest installment of “The Future of AI Credit” playbook, a PYMNTS Intelligence collaboration with Thredd, shows, creditworthiness is not necessarily constant. A borrower can be attractive for one transaction and unattractive for another.

A customer can have sufficient cash flow to support a $2,000 purchase on a six-month repayment schedule, but not a $2,000 revolving balance priced indefinitely. A merchant whose revenues fluctuate dramatically month to month may appear volatile under a conventional model even when its cash flows closely track predictable seasonal patterns.

The emerging opportunity is not simply to use artificial intelligence to make the old underwriting process faster. It is to change the unit that is itself being underwritten.

Using AI to Judge the Purchase, Not the Borrower

The report argues that the terms of a traditional credit product have historically been established largely independently of what the borrowed money will ultimately finance.

A lender can know whether someone paid a card bill late two years ago, how much revolving debt they carry and how long their oldest account has been open. What the traditional credit file does not see as well is the economic context surrounding the next dollar of credit: whether a small business is financing inventory likely to turn into revenue in 30 days, whether a gig worker’s irregular income is actually healthy cash flow or whether a particular purchase deserves the same price as every other purchase made by the same borrower.

That distinction is becoming more important as artificial intelligence collides with real-time transaction and cash flow data.

Instead of pricing the borrower once and forcing thousands of future purchases to inherit that judgment, lenders can increasingly evaluate the borrower, transaction and surrounding cash flow together. The result is potentially a credit market in which risk is priced closer to the moment it actually occurs — and customers who looked risky through the coarse lens of a credit file suddenly become financeable.

Read the report: Credit That Meets the Moment: What AI-Powered Credit Means for Customers

The consequential shift is from borrower-level underwriting toward transaction-level underwriting. That’s where AI credit begins to look less like a smarter version of today’s card and more like a different financial architecture.

Instead of asking, “What APR should this customer receive?” the system can ask, “What structure makes sense for this customer, for this purchase, right now?”

The answer might be revolving credit for one transaction, installments for another and debit for a third. Duration, repayment schedule and potentially price can become variables rather than permanent characteristics of the account.

The PYMNTS Intelligence report describes this as credit shifting from a fixed product toward a feature deployed when a transaction requires it, with real-time data allowing financing to reflect actual cash-flow behavior and transaction-level signals instead of relying exclusively on a generic risk tier.

AI’s credit opportunity may also be ultimately less about finding a cleverer way to reject risky borrowers than finding creditworthy borrowers hiding inside categories the old system labels uncertain. The underwriting question changes from “Does this borrower resemble historically safe borrowers?” to “What does the available evidence say about this borrower’s ability to support this particular obligation?”

At PYMNTS Intelligence, we work with businesses to uncover insights that fuel intelligent, data-driven discussions on changing customer expectations, a more connected economy and the strategic shifts necessary to achieve outcomes. With rigorous research methodologies and unwavering commitment to objective quality, we offer trusted data to grow your business. As our partner, you’ll have access to our diverse team of PhDs, researchers, data analysts, number crunchers, subject matter veterans and editorial experts.