He discussed five possible uses of machine learning in lending, which he prefaced by noting that before it is implemented at a credit union, the institution's data must be clean. "[AI] is a great tool, but it has qualifications," he said. "It has to have clean data, which is not easy for credit unions."

Automated underwriting. Credit unions could save a lot of money on costly third-party underwriting services and make more loans with fewer employees by having a machine do the underwriting for them. In addition, machines can often pinpoint decision-making factors that a human has never thought of. "Machine learning can help us connect the dots that we didn't even see could be connected," Hamilton said. "It automates, simplifies the process and makes new connections." He added that the intention of using machine learning in underwriting is not to eliminate jobs, but to take time-consuming, mundane tasks off employees' plates.
Credit line management. Many credit unions are already conducting research to see which members qualify for higher credit lines and following up with a targeted balance transfer campaign – machine learning can automate that process. "It's a good way to manage risk, increase the balances on cards, and encourage more use of the cards," he said.
Marketing. Credit union marketers already have access to tools that allow them to determine the "next best product" for a member – for example, when a member is likely to buy a car based on their previous car-purchasing frequency. But machine learning can take this a step further and call out unique indicators of certain purchasing behaviors, Hamilton said. A hypothetical example: A consumer who spends more than $300 a month in a pet supply store is more likely to be revolver on a credit card.
Collections. Machine learning can improve the collections process by prioritizing member calls. It monitors members' payment behaviors over time and determines which members should be called for payments first based on those behaviors, he said.
Loan processing. Using a product CU Direct is developing as an example, Hamilton demonstrated how a machine can autonomously process loan documents, dramatically reducing the amount of time credit union employees spend on manual loan processing tasks. He said CU Direct found credit unions that used machine learning processed 30% more loans compared to their production before implementing the tool.

Instant Insights /

5 Applications for Machine Learning in Lending: Onsite at REACH

Related Stories

Recommended For You