The recent spate of earnings reports from lending and insurance platforms offer up evidence that the loan underwriting via automation has expanded credit access and sharpened the identification of credit-worthy customers. The macro environment is fluid, and will prove a real test for credit metrics. But filings and other materials from the platforms reveal relatively strong performance to date.
Upstart’s ‘No Human Intervention’ Loans
Upstart, which operates an AI-underpinned lending platform, with banks and institutions on the lenders’ side extending personal loans to borrowers, said this month that 92% of its loans were fully automated, and in its earnings materials noted that with those loans there was “no human intervention by Upstart.” In the March quarter, the $2 billion in personal loan originations were up 83% year on year. In terms of volume, loan transactions across the platform were about 241,000, up 102% from the prior year.
The average loan size of about $8,865 was up from $8,580 in the prior quarter. Management noted on the call that the proportion of loans made to super prime borrowers increased. CEO Dave Girouard said the firm had seen “improved borrower health,” and said that higher conversion rates on lending helped boost revenues by 67%.
In terms of the modeling and AI, Girouard said new algorithms are successful at “clustering data that have meaningful relationships, allowing seemingly random data to become valuable to predicting credit performance.
“Our credit continues to perform well,” he said. “All else being equal, we believe a faster automated process selects for better borrowers.” The company’s latest quarterly SEC filing revealed that unrealized losses on loans and loan charge-offs were $3.8 million in the latest quarter, compared to $10.7 million in the previous year’s first quarter.
Lemonade, the insurance platform, said that in-force premiums were 27% higher in the first quarter than a year ago. In its earnings materials, the company noted that its vehicle insurance operations are growing, as “proprietary AI (LTV & telematics) pinpoints our target customer — young, safe drivers — with precision, offering unbeatable pricing. Industry-leading telematics adoption and a continuous flow of driving data enable us to accurately fine tune pricing to optimize underwriting performance.” Customer rosters increased by 21% to 2,545,496 as compared to the first quarter of 2024. Premiums per customer were $396 at the end of the first quarter, up 4% from a year ago.
LendingClub’s AI Push
LendingClub, which uses AI in risk assessment and loan underwriting, said at the end of last month that loan originations grow 21% in the first quarter to reach $2 billion. That same month, the company acquired AI firm Cushion. It will use the acquired firm’s technology to build on existing AI infrastructure to help users optimize payments, reduce interest, and improve credit health. Company filings reveal that net charge-offs in the most recent quarter were 4.8%, down from 6.9% a year ago.
Separately, and in its own quarterly results, Affirm noted that delinquency rates across several vintages of loans had improved year over year; for example, for fiscal year 2023, 30+ day delinquency rates were 2.3% in March compared to 2.7% in the September quarter.
As CEO Max Levchin said on the conference call with analysts, with a nod to advanced technologies, “We’ve used machine learning since the day of our founding. In fact, the idea was to build a credit score that was built on alternative data and modern machine learning techniques, and obviously we’ve been pretty successful with it.”
