Cursed with Knowledge? Strategic Risks of AI-Powered Monitoring

Abstract

Why are some financial institutions slow to adopt artificial intelligence (AI) despite its decreasing costs, while others are quick to embrace it? This paper develops a theoretical framework to understand the tradeoff between using AI for better borrower evaluation and maintaining strict financial discipline. Although AI improves the ability to assess borrower quality during refinancing, it can also encourage financiers to refinance initially speculative borrowers, thus exacerbating adverse selection in initial lending. To avoid these risks, some institutions may optimally choose not to adopt AI. Our model predicts that lower AI costs drive adoption only when this adverse selection is not severe. Moreover, AI adoption is most beneficial for institutions with a moderate-quality lending pool. These results help explain the strategic dynamics of AI adoption in lending and investment monitoring.

Keywords: Artificial Intelligence, Investment Monitoring, Strategic Information Acquisition, Soft Budget Constraint

JEL Classification: D82, G21, G23, G24, O32, O33