AI-driven embedded lending emerges as digital finance option
In this podcast, Yaacov Martin, CEO of embedded lending vendor Jifiti, explains how AI makes online credit efficient, accurate and profitable and why regulation fosters growth.
Buy now, pay later (BNPL) is an increasingly common option on retail payment terminals and e-commerce sites. It's part of the growing category of embedded finance that is changing how consumers and businesses access financial services – and how banks and merchants offer credit -- by making them available in non-financial applications.
Now AI is starting to transform BNPL and other types of online lending by handling the many steps of credit approval, including identity verification, decisioning -- analyzing creditworthiness and deciding whether to lend -- regulatory compliance checks and e-signatures. AI has the potential to speed up the information processing, analysis and decision-making and improve its accuracy. But it also carries some of the same risks of AI in other legally sensitive fields, such as talent management and medicine. For example, what happens if AI shows bias against minority groups in its credit decisions?
In this episode of Enterprise Apps Unpacked, Yaacov Martin, CEO and co-founder of Jifiti, explains how embedded finance platforms work, who uses them, and the benefits and challenges of automating the lending process with AI.
Jifiti makes a "white labeled" embedded lending platform used by banks, including Citi, Mastercard and Citizens, as well as retailers like Ikea and Peloton. It also offers a two-layer AI orchestration platform. The first layer uses agentic AI for loan discovery, in which lenders find potential borrowers and gather information, and loan origination, which encompasses the entire workflow from application to funds disbursement. The second layer employs a variety of AI methods for "core lending" processes, such as credit scoring, fraud detection and collections.
Better credit decisions through AI
Yaacov Martin
Martin explained what AI brings to the lending process while acknowledging risks that are holding back wider adoption.
"AI does a much better job at decisioning -- at creating a decision on behalf of the bank -- than any human underwriter can," he said. "Obviously, you need to train the AI, and you need to feed it the relevant data. If you feed it the same information that has been previously fed to different non-AI engines and human underwriters, you will get a better result." Lenders see fewer loan defaults as a result, he said.
But AI won't handle all the underwriting anytime soon because of laws against discrimination and the AI black box problem that prevents users from knowing how the technology arrives at its conclusions. "When you put all of this information into an AI, and the AI spits out an approval or decline, you have to ensure that you are still complying with all of those different regulations," Martin said. "Even those who are deeply familiar with the inner workings and underpinnings of AI will tell you that once the AI is trained, it is going to go ahead and produce the best result that it possibly can, sometimes trampling some of those principles."
But banks that shy away from AI are losing out on more than an efficient and profitable underwriting process, according to Martin. They could also lose business as more borrowers use generative AI to search online for credit and choose lenders that are already set up to respond quickly.
Other topics discussed in the podcast include the following:
- Where the Jifiti platform fits into the digital finance ecosystem.
- Why embedded lending might interest non-bank enterprises.
- What digital finance could look like in the future.
David Essex is an industry editor who creates in-depth content on enterprise applications, emerging technology and market trends for several Informa TechTarget websites.
