Inclusive lending

India’s banking system is becoming richer in data, but this is not yet translating into a wider borrower base. As Reserve Bank of India (RBI) Deputy Governor Shirish Chandra Murmu highlighted recently, the share of fresh businesses entering the formal credit system fell to 42 per cent in 2025-26, from 52 per cent in 2022-23, even as outstanding commercial credit grew 14 per cent over the year in 2025-26. Banks increasingly have access to goods and services tax filings, cash flows, utility payments, ecommerce records, mobile usage, and agricultural and geospatial data, yet traditional lending models remain dependent on collateral, financial statements, and formal credit histories. This leaves many viable small businesses and other “credit-invisible” borrowers outside the system. Alternative data and artificial intelligence (AI) can help bridge this gap. 

The next phase of financial inclusion must enable people across differences in language, location, livelihood, gender and access channels to participate meaningfully in formal finance. Economists have shown that information asymmetry can produce adverse selection and that markets can lose good opportunities when one side does not have enough information about the other. The same applies to lending. If a bank does not have enough information about a borrower, it should not automatically assume that the borrower is risky. This can deny credit to businesses that may have the ability to repay but lack a formal credit history. This is precisely where banks must leverage their technological capabilities. India already has examples of this approach. Fintech lenders have used transaction data from small retailers and platform workers to assess repayment capacity. The World Bank, for instance, has documented how State Bank of India’s e-Smart SME project used digital information to provide collateral-free working-capital loans to ecommerce sellers. Such models could become more scalable as India’s digital public infrastructure and consent-based data-sharing architecture deepen. 

The way forward, therefore, is using data, models and technology alongside human intelligence. Experience, knowledge and judgement remain essential to assessing whether a model’s output makes sense and to intervene when it does not. Governance intelligence is equally important, with boards and senior management responsible for understanding the risks and limits of AI systems. Finally, collective intelligence can allow information to be pooled across the financial system, helping lenders identify patterns and assess risks more effectively. Banks must also ensure that alternative data is used with consent and subject to regular oversight. They should regularly test models for bias and reliability, particularly as they are extended to new customer groups, and ensure that customers can seek a meaningful human review of decisions that materially affect them. Unreliable data, algorithmic bias, privacy concerns and opaque models could create new exclusions. The test of technological progress will ultimately be whether it helps banks reach viable borrowers they could not serve earlier.