Lenders weigh strengths, risks of AI-driven loan process

Equipment financiers are adopting AI-driven technology though potential biases in the algorithms present concerns. 

As equipment lenders increase their use of AI for workflow automation, fraud detection and new financial models, borrowers could benefit from a more personalized loan application process, Oakmont Capital Services Chief Operating Officer Daryn Lecy told Equipment Finance News this week during the Equipment Leasing and Finance Association convention in Austin, Texas.  

“I think enhancing the borrower experience could be from the [loan] application standpoint, where there could be more of a Q and A with a chatbot or an individual AI assistant,” he said. “With time, AI will know enough about you — hopefully the industries that we work in and the equipment that we’re very educated in — to ask concise questions that are to the point.” 

Roughly 72% of lending institutions use AI, up from 55% in 2023, according to software provider JumpGrowth. 

Personalization is one of AI’s main strengths, Scott Nelson, president and chief technology officer at St. Paul, Minn.-based Tamarack Technology, an equipment finance software provider, told EFN during the event. 

“If customers want to be treated differently, you can write systems that will personalize pretty fast,” he said. “AI will be good at that.” 

Streamlining the process vs. biases in AI 

AI can enhance customer service by streamlining the lending process through automation and self-service tools, Patricio Pazmino, head of analytics and AI at Quito, Ecuador-based Kin Analytics, told EFN. 

“It would be great for borrowers to do the whole application process through digital channels, but even now there are still a lot of manual channels with hard-copy documentation,” he said. “That brings a lot of time into the process, and through AI, you can solve that. You can automate all that document extraction of data” to provide quick responses. 

While AI can evaluate borrowers’ creditworthiness to make smarter loans more quickly, it also may perpetuate existing biases in the data that AI systems are trained on. Thus, some borrowers could be denied when they should be approved based on the lending institution’s criteria, Tamarack’s Nelson said.  

In one instance, an AI system denied an entire pool of applicants because their time in business was two years or less, despite analysts predicting that they would be approved, Nelson said. It was later discovered that in the AI database, 20% of loans to borrowers with less than two or three years in business were delinquent. 

“We identified a policy that could be looked at in terms of ‘we could do more,’” he said. “But think about it, it’s hard to ask an AI model with 60 variables why you were declined.”