Banking – loan application processing

Lorykeet enables multiple testing methods, detects potential bias and discrimination in credit risk algorithms, and provides counterfactual capabilities to help determine options for growing loan portfolio within risk parameters

The use of machine learning/artificial intelligence in loan processing is quickly becoming standard practice. Machine learning determines factors relevant to credit risk and loan approval decisions. AI can complete credit and background research without human supervision. The algorithms learn how people have behaved in the past, replicate that behavior, and predict whether a loan applicant is a good credit risk. Using machine learning in loan processing and decision-making speeds the process and reduces errors.  Using AI in this manner comes with compliance risks. Lenders need to ensure that they do not violate any laws, rules, regulations, or internal policies.

Lorykeet’s platform ingested the data and algorithms produced by the ML/AI and provided explanations (XAI) for the outcomes and predictions produced by the analyses. Lorykeet provided testing capabilities, including methods typically used by data scientists, risk managers, and auditors, leveraging its “human-in-the-loop” design.  Lorykeet’s counterfactual capabilities allowed determination of how applicants who were not approved could achieve approval, and how the lender could adjust certainly factors or metrics to enable approval.  At the macro-level, Lorykeet helped to determine how the lender could increase the percentage of loan applications approved without significantly increasing its credit risk.

Process automation, machine learning, and AI are now embedded in the loan processing sector. Their use will clearly expand in the future. This revolution in the application of AI comes with new risks that need to be managed. Lorykeet’s ability to enable understanding, explainability, and assurance regarding trustworthiness and compliance will grow in importance as AI takes on an ever-increasing role in loan processing. 

Because certain steps in loan processing can be accomplished with AI does not mean that the need human involvement is eliminated. Lorykeet’s human-in-the-loop capabilities will allow different stakeholders to do their part in managing risks, assuring compliance, and improving decision-making. Loan applicants may still want humans to talk with about the process and the results (at least in the near term). And with AI able to consider thousands of variables, lenders will need tools that allow them to understand and explain why their algorithms are producing the outcomes and predictions they are. 

Comments are closed.