AI-Enabled Credit Scoring and Financial Inclusion in Indian MSME Lending: An Empirical Analysis of Risk Assessment, Loan Accessibility, and Borrower Performance

Abstract

The use of AI in credit scoring will influence risk assessment, loan sanctioning and borrower performance in MSME lending in India. The research addresses a central issue in MSME financing. The small-sized firms often need timely working-capital credit. However, many of them do not have collateral. A few have undergone audit in the past. A long credit history and standardised documentation are also absent in many of them. An artificial empirical dataset was created for educational purposes with 300 observations of MSME borrowers and 50 observations of lending-professionals from banks, NBFC, fintech lenders, and MFI. The simulated design management includes elements like sectors, geographical areas, borrower inclusion status, alternative-data usage, loan approval, processing time, collateral requirement, perception of transparency, satisfaction of borrowers, repayment orientation and business performance. The study utilized reliability analysis, descriptive statistics, correlation analysis, and regression analyses, accompanied by t-test and ANOVA tests. The evidence presented suggests improved loan outcomes when lenders utilize AI-enabled channels as well as alternative-data-based channels. Specifically, using AI-enabled channels is associated with a higher probability of loan approval and shorter loan-processing time. Furthermore, the use of these channels are correlated with lower collateral dependence and better borrower performance indicators. Transparency and trust are crucial because otherwise, the automatic credit decisions may reproduce informational and social exclusion. The paper presents an empirical framework focused on borrower insights to study AI in micro, small and medium enterprises (MSME). Digital lending that is inclusive requires explainability, consent-based data use, model governance grievance redressal and continual bias monitoring. The findings should be seen as an analytical demonstration as opposed to field evidence, which requires primary data and lender-level verification to deploy.