5 Myths About AI from Small- and Mid-Sized Lenders

Almost every industry uses analytics these days. Technological advances have helped companies of all sizes become more targeted in sales and marketing and more efficient in their operations. But the rapid pace of change can leave some small- and mid-sized lenders reluctant to implement analytics and unsure how it might benefit them.

I talked about this issue at American Banker’s Retail Banking 2019 conference in Austin, Texas. Joining me was Nelly Rojas-Moreno, chief credit officer for LiftFund. LiftFund, a nonprofit Community Development Financial Institution (CDFI), partnered with 2River Consulting Group. We transformed their historical data on loans to entrepreneurs in low-to-moderate-income areas into an automated, AI-based decisioning model. This investment has allowed LiftFund to grow loan applications and loan volume while also improving processing time (including next day approvals) and improving repayment rates.

LiftFund’s success got me thinking about what’s holding back small lenders from reaping the benefits of analytics. Here are five myths you may have heard about analytics and why they’re not true:

1. I don’t have enough data.

Some low-volume lenders think they need decades of records to train AI models. But most mid-sized financial services companies have looked at enough deals to have the 3,000 or so historical records needed for a model. And even if they don’t have enough lending records for AI, smaller lenders can implement analytics in other parts of the business. For example, if your company does any email marketing, chances are you’ve gathered enough data to use analytics to refine your messaging strategy.

2. I don’t have the right people.

Many lenders think they need data science teams to get the most out of analytics. While that might help in setting up the initial model, your underwriting team is probably already analytical enough to understand and effectively work with AI models. That’s one of the reasons our analysts and data scientists love working with underwriters. If your underwriting team is using a scorecard to rate lending risks, they’ll be comfortable working with an analytics model.

3. Lending is more of an art than a science.

This one isn’t really a myth. But it’s only one part of the story. Lenders, especially mission-based ones like LiftFund, take pride in their personalized service and their ability to identify good risks who might be rejected by traditional metrics. Borrowers are real people, and so are underwriters. Analytics don’t replace relationships. In fact, implementing analytics models can help free up time spent gathering data, allowing underwriters to focus on personalization.

4. My credit processes don’t allow modeling.

Most lenders have credit committees that require easily understandable, step-by-step documentation of credit decisions. For these lenders, “black box” AI models might not be the best place to start. But analytics can help bring large data sets together to create sophisticated decision trees that are easy to understand and will flag potential risk factors before they happen.

5. I’m not digital.

Some smaller lenders have only recently moved away from paper-based record-keeping systems. Just digitizing the boxes of paper records can seem like an unsurmountable task, let alone implementing any AI model. But lower-tech lenders should consider analytics as part of their broader digitization strategy. If you’re using Excel to manage your existing portfolio, you’re can use analytics.

Analytics can help even the smallest of lenders better understand their customer base and process higher volumes without having to hire new staff. It can also free up time to focus on the personal relationships in which many lenders take pride.

Recent Posts

The Archive