AI Is Changing How Credit Decisions Are Made
By CU Today Staff —
NEW YORK--Credit decisions are increasingly being driven by consumers' real-time financial behavior rather than traditional static credit scores, a shift that promises more accurate underwriting while reducing fraud and false declines, according to PYMNTS.
PYMNTS reported that lenders are moving beyond one-time credit decisions by analyzing current cash flow, spending patterns and transaction histories. The shift gained momentum in June when Plaid introduced a foundation model that evaluates the sequence and timing of a consumer's transaction history instead of relying primarily on a single credit score. According to the company, early testing showed the model reduced default risk by 13.6% while cutting returned-payment losses by 26.5%.
The move comes as merchants continue to struggle with balancing fraud prevention and customer experience. Citing its latest fraud prevention tracker, PYMNTS Intelligence reported that overly aggressive fraud controls cost merchants an estimated $50 billion annually, with nearly half saying as many as 5% of legitimate transactions are mistakenly rejected. The report said AI models that evaluate behavioral patterns instead of static data can improve approval rates while reducing fraud losses and costly false declines.
The trend is also spreading into business lending. PYMNTS noted that Billtrust recently launched its Agentic Credit Lines platform, which uses payment data from 13 million business buyers to recommend commercial credit limits before payment problems develop. Meanwhile, Deutsche Bank Chief Risk Officer Marcus Chromik told McKinsey that credit reviews that once required days can now be completed in near real time, allowing financial institutions to make faster lending decisions.
According to PYMNTS, the next evolution is "agentic credit," in which every transaction is evaluated individually rather than applying the same credit decision to every purchase. Affirm President Libor Michalek told PYMNTS that the approach considers factors such as the size of a purchase, a borrower's current debt obligations and monthly cash flow before approving financing. The report concluded that while AI-powered underwriting offers significant gains in risk management and member experience, financial institutions will need modern, connected data infrastructure to deliver real-time credit intelligence at scale.
Originally reported by CU Today.