Banks and credit unions see AI-powered software tools as key to offering more tailored customer service and breaking down old ways of making loans that unfairly wrote off potential customers. iStock illustration

A year after ChatGPT burst onto the national stage, Massachusetts’ banks and credit unions have begun adopting artificial intelligence-powered software tools in quests to enhance customer experience, offer new products or improve operational efficiency.

Credit unions and their bank peers are under pressure from “the Amazon effect,” said Traci Michel, chief operating officer of Chelsea-based Metro Credit Union, and need to find ways to offer more personalized service in digital channels.

“When a consumer is having these hyper-personalized and streamlined experiences in one aspect of their life, they apply those expectations elsewhere – outside of the industry where it originated. The rewards of using AI across financial services are significant: elevated customer service, improved employee experience and increased efficiency. AI is a great tool to empower members and make faster, more accurate support a reality,” she said.

Michel said the credit union is using generative AI – the same type of software behind ChatGPT – to power a chatbot that helps members with basic banking needs such as getting account or loan balances. She said this will also improve Metro’s understanding of customer needs and help it personalize how it gives them the information they need.

“Financial services is a data-driven industry and a natural fit for AI,” Michel said. “Since 2022, Metro’s digital assistant, ‘Chelsea,’ has chatted via text with members online and recently expanded her duties to answer questions via phone, while live Metro agents are always available to assist with more complicated requests.”

She said the Chelsea chatbot, built with software from Posh AI, reduces the time it takes to give members answers to their questions, allows 24/7 access to information from any location, and adds a more rewarding experience.

In the same way, Michel said employees can also turn to AI to access knowledge or information that they need to answer basic questions and provide them with quick and accurate information to resolve member issues and questions.

With these basic functions covered, she noted that employees can use the time thus freed up to focus on more difficult and complex requests like opening new accounts or applying for loans.

“Consumers want and expect their financial institutions to meet their expectations for fast and easy access to what they need,” Michel said. “To keep up with the competition and evolve with the latest technologies, Metro and other community financial organizations must adapt to these changes or risk being left behind.”

Better Small Business Underwriting

Artificial intelligence software can also be used in automating aspects of small business loan underwriting, while helping lenders increase loan approvals for minority business owners and first-time entrepreneurs, said Patrick Reily, co-founder of Uplinq, a software company helping lenders with data and scoring solutions.

Reily said AI is capable of processing vast amounts of data and producing reports and recommendations on which borrowers are credible and are likely to pay obligations.

He said the key is to sift through massive data that can give better context to each borrower’s situation. He said the more data being fed into an AI-powered piece of software, the better it can identify what makes a small business owner or other borrower special.

This can include data banks own, such as information found on loan applications, or external data like credit scores, as well as other types of information that can tell the complete view of the business owner and the business, the larger economy, the competitive landscape in the applicant’s industry or the market for their product or service, which can improve loan approvals and give more access to capital to those who need it.

“AI can look across an ocean of information and describe success factors [for borrowers]. That’s one of the ways we’re able to say ‘yes’ more often to people who are traditionally declined [after submitting loan applications],” Reily said. “[Beyond credit scores, AI can] look at people in an equal way, see their creativity, ingenuity and ability to solve problems. We can still achieve the same kind of risk performance but we’re able to reach populations that were historically underserved because they don’t fit a cookie-cutter mold of approval.”

Declining to name the bank, he said Uplinq helped a client – a Central Massachusetts bank – improve its underwriting and lending to minority business owners in the communities it served.

“We looked at how effectively they were reaching target CRA communities, the traditionally underserved segments of the community. We were more inclusive in the information [that] we were taking, and we crafted more precise underwriting models,” Reily said. “This led to increasing loan approvals from the minority communities from 25 percent to 65 percent approval, and this included women, people of color, immigrants, first-time entrepreneurs and gig workers.”

He said this represented an improvement from traditional underwriting approaches where small businesses first need to establish a steady stream of income or have hard assets like real estate that can secure a loan before gaining access to capital.

More than Fraud Prevention

AI-powered fraud prevention tools are not new in the banking industry. But AI technology has also begun making its way into a number of other uses, from process automation to predictive messaging used in collections and recovery workflows, Metro Credit Union’s Michel said.

AI can also power tools intended to help customers better manage their finances. Boston-based OneUnited Bank launched in October last year its AI-powered WiseOne Insights, which provides real-time guidance and notifications to help its customers make better financial decisions.

Nika Cataldo

Being the biggest Black-owned bank in the US, OneUnited Bank said it aims to close the racial wealth gap by having the AI tool notify its customers timely on bill payments, advise them when extra funds are available for savings, recommend personalized debt payoff strategies, and other actionable insights.

Kevin Cohee, chairman and CEO of OneUnited Bank, said the AI tool will serve as a financial wellness companion to customers, addressing the need for improved financial literacy.

“It is about building a financial wellness ecosystem. Using technology, we can build these self-supporting systems to empower all,” Cohee said.

Michel said Metro is also exploring other additional use cases with technology providers that specializing in different types of AI applications, which are either in a proof-of-concept or exploratory stage.

Not ‘Set It and Forget It’

But despite the positive buzz artificial intelligence has gained in many corners of the financial services industry, a recent national J.D. Power survey found that most customers of credit unions and small banks are less familiar with the technology, and many are skeptical about how the technology, when applied to banking, can make their lives better.

Around 80 percent of consumers in the survey think the increasing use of AI in financial services puts them at “somewhat” or “extremely” greater risk for fraud and security breaches.

Between one-sixth and one-quarter of the American consumers polled by J.D. Power said they would never use AI for a range of tasks, from making investment recommendations (14 percent) to using facial recognition technology to withdraw cash at an ATM (26 percent), with large shares of consumers wanting to see how AI-powered services play out before using them. Notably, 18 percent said they would never use an AI-powered chatbot for customer service.

Currently, Metro Credit Union is making sure its AI tools are being overseen by the lender’s staff, Michel said.

“Behind all of our AI applications are real people – employees with extensive Metro experience. They are watching transcripts of AI conversations, listening to interactions in near-real time, and making continuous improvements to workflows to better the service experience on a daily basis,” she said. “This isn’t ‘set it and forget it’ – nor should it be for any organization leveraging this type of technology.”