Artificial Intelligence in Credit Risk Management: Would you use it?

The global pandemic since 2020 is a rare occurrence with significant consequences. Such events necessitate novel technology to assist lenders in balancing portfolio risk while providing a favorable borrower experience. LendIt and Brighterion conducted a survey among financial institutions (list not shared) to know how they are likely to use technology such as artificial intelligence (AI) in credit risk management given the effects of the pandemic. in an article, Tom Bittlemann, Senior Specialist, Product Development at Mastercard summarized five significant takeaways from the study.

Among the financial institutions and lenders (FIs) that participated in the survey, 50% are currently using AI for credit risk management, with another 25% planning to do so in the future. About 88% of respondents plan to boost their AI investment In the next two to five years. These results are expected according to Bittlemann because AI is providing significant benefits to Brighterion’s loan customers. These benefits include improved customer experience, earlier risk detection, and annual savings in the tens of millions of dollars.

RULES-BASED SYSTEMS vs. AI

“Rules-based systems” ranks among the top credit risk management technologies utilized today whereas AI/machine learning is the second. Bittleman says that rules-based systems lack adaptability especially during the unpredictable times of pandemic because you cannot catch every guideline for credit risk management. Human-designed credit risk management systems with the assistance of AI models will make the process resilient. Data revisions of the system’s current model along with the recent laws and guidelines are applied to determine the best adjustment for the company and client.

REAL-TIME SCORING USING AI

Bittleman believes that data from credit bureaus are only a good lagging indicator of prior client behavior but does not provide a clear picture for current behavior. Real-time risk scoring is an advantage of using AI. While historical patterns can help put economic data into perspective, only real-time scoring can look at particular trends as they occur.

MODELING IN AI

Around 32% of financial FIs fear that AI is difficult to understand and explain. Given the potential for adverse outcomes as AI usage grows, model explainability has become increasingly important. Bittlemann says that building complex models is not better than accurate and explainable models.

About 26% of the survey respondents said that difficulty of AI implementation is among the most important concerns. Developing an AI model for businesses takes months, if not years. Many people have tried and failed to do it in-house. This led to many people believing that implementing AI would be difficult.

USING AI IN THE FUTURE

There is diversity among the respondents regarding how they would use AI for credit risk in the future – 30% would utilize AI for new client loan origination, 30% for credit risk monitoring, and 27% for collection optimization. The remaining 13% stated “other uses” or “none of the above.”

AI STRENGTH

Lenders want to know how likely a client is to default on a loan and how this affects profitability, what credit limit should be allocated to a consumer and when to act, and how to optimize collection techniques with these aspects in mind. Risk considerations, according to Bittlemann, must be evaluated alongside revenue drivers such as interest income, interchange revenue, and fees. That’s where AI comes in handy, as it can help with important decisions throughout the customer lifecycle, ultimately strengthening the customer connection.

Reference: Brighterion