An analysis said trust in decisions is becoming as important as prediction accuracy in the financial sector's AI loan screening competition. As more financial companies use similar technology, the ability to explain loan outcomes consistently and fairly is emerging as a differentiator.

Fintech Futures reported on Sept. 17 that Subramanian Narayanaswamy (수브라마니안 나라야나스와미), a managing director at Wells Fargo, said in a contributed article that major lenders are using similar credit-scoring data and machine-learning technology. He stressed the importance of building trust in AI models.

Narayanaswamy presented four conditions for trustworthy lending AI. It should produce consistent results for applicants in similar circumstances and explain loan denials in a way customers can understand. It is also important to verify that specific borrower groups are not disadvantaged without reasonable grounds and to continuously check model performance as economic conditions or customer behavior changes.

Trust can also affect profitability and operational efficiency. For example, even for self-employed borrowers whose monthly income is irregular, there is room to assess repayment ability more accurately by also reviewing annual cash flow. Clear grounds for decisions can also help financial companies with model validation and audit responses.

The importance is also growing of materials that demonstrate explainability and fairness when financial companies adopt external AI solutions. Zest AI said it operates more than 600 lending models and has supported loans worth $1 trillion, or about 1,381 trillion won, while expanding approvals without increasing credit losses. FairPlay checks for unfair outcomes between borrower groups, and Credo AI supports model monitoring and audit material management.

Narayanaswamy said the competitiveness of lending AI will depend not only on model sophistication but also on the ability to continually demonstrate accuracy, fairness and explainability.