Aye Finance has completed a pilot of a generative artificial intelligence model that estimates business sales from images of store premises to enhance formal credit for trading businesses in tier two and beyond cities. The in-house solution combines generative AI and machine learning to convert unstructured visual data into financial inputs and to lower the cost to serve micro enterprise customers. The pilot focuses on garment and grocery stores and aims to expand access to formal lending for underserved micro MSMEs.

The system uses a multimodal large language model (MLLM) integrated with proprietary machine learning models to provide reliable estimates of monthly sales from store images together with other parameters. Development drew on extensive internal datasets and testing to ensure consistency and to reduce reliance on subjective individual judgement in income assessment. Automation of these components enables faster processing and more standardised underwriting decisions.

Aye Finance indicated the model will substantially reduce manual fieldwork and improve the speed and reliability of credit decisions as part of broader efforts to optimise operational efficiency. The company established a data science and artificial intelligence unit in 2019 and has deployed multiple customised machine learning models across the customer lifecycle to improve unit economics. The approach opens the possibility of applying the methodology beyond trading to other sectors and industries.

The initiative targets a critical barrier to credit for about 60 million (mn) micro entrepreneurs in India's tier two cities and beyond who lack formal accounting records and paper trails. As a middle-layer non-banking financial company, Aye Finance has offered tailored working capital and expansion loans since 2014, typically secured against working assets or property. The company recently closed an initial public offering that raised Rs 10,100 mn to strengthen its capital base.