Aadhaar and UPI have digitised Indian banking, but AI-driven continuous underwriting remains rare. Economics may explain part of it, but there appears to be more behind the slow adoption. TransUnion CIBIL puts India’s credit eligible population at 1.036 billion. Only 277 million actively use formal credit. The remaining 759 million either do not need credit, are bad agents, or have limited data available to bureau-based underwriting because it scores borrowing history, not financial behaviour.
Snapshots to Real-Time Intelligence
Most lending still relies on bureau scores, income proof and repayment history. This model retrained once a quarter describes a borrower who no longer exists by the time a loan is priced. Underwriting, KYC and fraud collection stacks operate separately, limiting real-time insights. Integrating these pipelines would enable continuous underwriting, where every transaction updates a borrower’s risk profile instead of waiting for the next credit cycle.
Consider a precision-engineering supplier with a clean repayment record. Suppose a large customer began delaying payments, supplier’s bank balance thins. Yet every EMI is still paid on time, funded from reserves. A quarterly model sees a compliant account and a healthy borrower, while continuous underwriting detects emerging stress early, enabling lenders to restructure the loan before default occurs.
The same intelligence applies to the underserved. A direct-to-consumer apparel founder with no bureau history but with rising digital revenue, regular GST filings and disciplined supplier payments is unfundable on a credit score alone. Recently, a unicorn founder was rejected by a bank because he was considered risky! By analysing real-time cash flows instead of past borrowing, lenders can assess such businesses earlier. The RBI’s move to weekly credit score updates from April 2026 signals this shift.
AI-native credit infrastructure
Gross NPAs of scheduled commercial banks stood at a multi-decadal low of 1.8 per cent as of March 2026, according to the RBI’s Financial Stability Report, reflecting a benign credit cycle rather than better monitoring. It also flagged AI-enabled cyberattacks as the top risk cited by banks and NBFCs themselves. Continuous financial data proportionally expands underwriting engines’ cyberattack potential. So, governance cannot be an afterthought. FREE-AI says banks remain responsible for AI-based lending decisions. Borrowers must be told when AI is used and have the right to challenge decisions. Combined with data protection and consent rules, this makes continuous underwriting more transparent, accountable and fair. This can be called AI-native credit infrastructure as it will runs on top every function, from onboarding to collections, working from a shared, continuously updated model of borrower behaviour, governed by consent that is obtained rather than assumed by default.
