Don’t get caught in the lending slow lane
Three ways to drive lending efficiency with AI
The most successful businesses strive for and achieve operational efficiency to continue their upward momentum. Whether it’s prioritizing key initiatives, minimizing wasted resources, or automating processes, efficiency has a positive ripple effect on everything from the bottom line to employee morale.
It can feel overwhelming to create and sustain efficiency, which can push some to cut corners. In the financial services and lending industry, cutting corners for the sake of efficiency is a dangerous possibility that can lead to legal liability, broken trust, and damage to the reputation of an organization or even the entire industry.
To better understand their lending processes Cornerstone Advisors conducted a study with leaders of community-based financial institutions (credit unions and community banks) across the U.S. Not surprisingly, according to the report, efficiency is one of the top concerns in 2024.
The good news is that lenders don’t have to sacrifice caution for efficiency (or vice versa). Cornerstone’s research shows there are opportunities to increase efficiency with credit modeling and data, auto-decisioning, and the adoption of AI technology in lending.
Credit modeling and data
The foundation for efficiency in lending is a well-built and maintained credit model. One of the most important attributes of a strong credit model is often overlooked: lender confidence. A surprisingly low number of institutions (13 percent) would say they and their credit models are “very prepared” for changing market conditions. This is echoed in the 63 percent who indicated it’s “somewhat difficult” or “very difficult” to update their credit models.
Optimizing for confidence in the model’s output is more obvious than you think and starts by using validated, purposeful, compliant data sets and proven outputs.
Data scientists use the expression “garbage in, garbage out” to concisely illustrate what happens when bad data is used. The same idea applies to credit models. Using reliable data sets with performant models creates more accurate credit decisions that financial institutions can feel confident using and updating if needed with changing economic conditions. This is even more important to take into account as more financial institutions look to include more variety in their data sources.
According to Cornerstone’s report, more community-based financial institutions are using nontraditional data sources. In fact, two-thirds reported using at least one nontraditional source in their credit modeling. The top three most common nontraditional sources were employment history, deposit account transaction patterns, and FCRA-compliant alternative data.
At the end of the day, when an institution can’t trust the decisions of a credit model its operational efficiency will decline. It’s important to be critical of what is being used to generate decisions and ensure that the model is performant, compliant, and transparent. The consequences of not doing this are more manual reviews and playing catch-up to the changing market.
Auto-decisioning
Roughly eight out of 10 community-based financial institutions have auto-decisioning systems in place and the grass appears to be greener for them. However, only one-third of the surveyed institutions are auto-decisioning more than 40 percent or more of their business lines.
Manual reviews are still common even with auto-decisioning. Almost half of the respondents say they manually review up to 25 percent of their auto-decisioned loans. One out of five will spend an average of 30-60 minutes on a manual review.
Overall, financial institutions still see a significant increase in efficiency through automated decisioning. Cornerstone Advisors benchmark data indicates “the number of loan applications reviewed per underwriting full-time equivalent (FTE) employee per month is 3.5 times greater among institutions that use automated decisioning than those that don’t.”
Using AI technology in lending
Most new technology faces resistance before it is embraced commercially. The same is true for AI and the lending industry. When it comes to whether or not AI technologies can and will improve the credit and lending process, 80 percent of respondents agreed, but only 13 percent have actually adopted the technology.
The reasons for being hesitant vary by institution but a top concern for most is regulatory compliance. Other concerns like “Cost to acquire,” “Transparency,” “Lack of knowledge,” and “Bias” all rank fairly evenly as the next most common concerns.
Even still, these concerns are not enough to keep community-based institutions from taking the next steps toward adopting AI technologies in the near future. Over half of these institutions plan to invest $100,000 or more into AI-driven models in the next three years.
Financial institutions can start the path toward lending efficiency now which will allow them to thrive in the future. The institutions that embrace more advanced technology, like AI, will see their investments pay off with stronger performance tied to more efficient lending operations. It’s only a matter of time before those who hold out will be the minority.
If your organization is considering taking the next step to future-proofing your lending efficiencies, ask these questions:
- Do you have the right resources and partnerships to develop effective and accurate credit models?
- Could automated decisioning systems help scale your lending operations safely and efficiently?
- Are you investing in enough innovative technologies, like AI, to boost your automation and competitive advantage?
