Fintech Deploys GenAI for Underwriting

Aye Finance's model employs a Multimodal Large Language Model (MLLM) to analyze unstructured visual data. This system processes images of storefronts, inventory, and other physical parameters to generate estimates of monthly sales for businesses like garment and grocery stores, which often lack formal financial records. Founded in 2014, the Gurgaon-based lender has a history of tech-focused backing, notably being the first Indian NBFC to receive an equity investment from Google's CapitalG in 2018. This investment preceded the 2019 establishment of their dedicated Data Science and Artificial Intelligence Unit, which has since deployed multiple custom ML models across the loan lifecycle. The core problem this technology addresses is the credit gap for India's 60 million micro-MSMEs, particularly in Tier 2 cities and beyond. These businesses often cannot provide the formal accounting documents and paper trails required by traditional underwriting processes, making it difficult to access formal credit. By building and training the model on its extensive internal datasets, Aye Finance aims to standardize a previously subjective evaluation process. The system is designed to create uniform, consistent, and faster underwriting decisions, significantly reducing the "cost-to-serve" for this micro-enterprise segment. Since its inception, Aye Finance has disbursed over ₹8,000 crores (approximately $960 million USD) to more than 700,000 unorganized businesses. The company recently raised INR 1,010 crores in an IPO and secured $37 million in a Series F funding round in late 2023, signaling strong investor confidence in its tech-driven approach to financial inclusion. This application of computer vision is part of a broader fintech trend to automate and improve risk assessment. Similar technologies are used for document verification (KYC), real estate valuation from satellite imagery, and assessing insurance claims from photos of damaged assets, replacing time-consuming manual inspections.