With new research finding racial disparities in small business lending, one fintech is examining how artificial intelligence (AI) and automation can help improve fairness and reduce inefficiency that may be present in traditional lending processes.  

Bank Automation News spoke with Sandip Nayak, chief strategy and artificial intelligence officer at Linear Financial Technologies, a digital account origination platform used primarily for small and midsize business (SMB) programs.  

Reston, Va.-based Linear counts among its customers $186.1 billion Citizens Financial Group, $203.9 billion Fifth Third Bank and $554.8 billion PNC. The fintech was founded in February with the merger of embedded credit solutions provider Fundation and small business lending solutions firm ODX Solutions.  

Nayak noted that banks and financial institutions, for the most part, are grounded in principles of responsible lending and delivering fairness to their customers, but problems and bias exist nonetheless, as suggested by research from New York University and automation platform provider Ocrolus.   

“Things can be improved and made much more efficient,” he said. “While there may not be intentional consequences by design, and in spite of responsible practices, there could be unintended consequences, for whatever reason.”  

“Using the latest technology — [AI, machine learning] and automation — you have to have some guiding principles, but that can certainly help in dealing with these socioeconomic and fairness issues,” added Nayak.  

AI, machine learning (ML) and automation can help root out bias and other problems in lending by:  

1. Increasing predictive power

AI can boost predictive power in lending decisions, resulting in better pricing and better products for customers, Nayak said.  

2. Increasing approval rates  

Lenders “can also be more inclusive — you can increase approval rates,” Nayak noted, “because of the ability of AI to be more productive.”  

3. Eliminating manual processes 

ML and automation can eliminate manual processes in business lending, a probable source of inefficiency and any bias that may be present. “Biases may not be intentional, but [manual processes] can definitely be a source of decisions which may not be responsible,” Nayak said.  

4. Right-sizing loan products 

Pulling in alternative data can improve the lending process as well. “Bank-level transaction data, which we do use … can definitely add a layer of ability to pay assessments,” he pointed out. “This is one of our core tenets where we are not overextending the customer.”  

5. Improving fraud detection 

“This is not spoken of in the context of responsible finance, but there is advanced fraud detection” possible with AI and automation, said Nayak, and eliminating more financial crime perpetrators and fraudsters allows more financing for good players. Higher fraud equals more inefficiency for legitimate customers.  

In addition, AI and automation stand to improve lengthy manual processes traditionally present in SMB lending. “The way banks have done this is to use commercial analysis,” Nayak explained.  

This analysis can include “heavy manual processing” of things such as tax returns, financial statements, income statements, balance sheets and more. It’s an antiquated process particularly in the case of small businesses, which make up much of the underserved category, Nayak told BAN.  

By contrast, “using AI and automation to serve the small businesses is very customer-centric,” because a small business owner can thereby get a loan or line of credit in a few days as opposed to weeks or months.  

Ruling out rules

Many banks still use rules-based systems in lending, where lenders set parameters or rules to evaluate customers’ creditworthiness, Nayak said, noting that this can lead to suboptimal credit decisions for SMBs. “There are all kinds of small businesses — different industries, different corporate structure, different sizes, young, mature, and so on,” he noted.  

“Rules-based systems, which have been the way for a long, long time, are not only suboptimal but inefficient,” Nayak said. Instead, he suggested that systems powered by AI and automation can better serve the “very heterogenous” small business segment.  

The upshot? AI and automation can indeed help level the playing field in lending — if the automation itself is fair.    

“It’s tremendous, and I’m not using that lightly,” Nayak said of automation and AI’s potential in this regard. “There can definitely be significant benefits. But the caveat is, there has to be a sense of responsibility of AI as well.”