By Joydip Gupta, APAC Head, Scienaptic AI
Indian banks just posted their highest loan loss provisions in three years, ₹28,883.6 crore in the June quarter¹, while simultaneously maintaining some of the cleanest balance sheets they’ve had in decades. Slippages jumped 26% year-on-year to around ₹63,000 crore, yet the system’s gross NPA ratio sits pretty at around 2.3%³, near multi-decade lows and expected to stay there through FY26.
It’s like watching someone buy an expensive umbrella after getting soaked in the rain, then bragging about staying dry while standing indoors. Panic? Or preparation? Maybe they are not reacting to today’s problems; they’re preparing for tomorrow’s possibilities.
But here’s the real question: instead of just building better umbrellas, shouldn’t they be building a more accurate weather forecast?
The New Credit Reality
Remember that friend who suddenly became health-conscious after turning 40? Indian banks are having their wellness moment. After years of aggressive growth in unsecured retail lending and small business credit, reality is setting in. The hangover from two years of rapid expansion is showing up in early delinquency buckets, particularly among borrowers who discovered that having five credit cards doesn’t actually multiply your ability to pay by five.
Private banks, with their hefty personal loan and credit card books, felt this shift first. It’s not that their customers suddenly became irresponsible overnight. Rather, as the refinancing party wound down and interest rates stayed elevated, marginal borrowers found themselves in that uncomfortable space between “I can totally afford this” and “Oh wait, can I actually afford this?”
Meanwhile, public sector banks have been quietly playing the tortoise to everyone else’s hare. They gained market share in the boring-but-profitable home loan segment⁵ while private banks wrestled with their unsecured exposures. Sometimes, slow and steady doesn’t just win the race; it also helps you avoid explaining embarrassing losses to shareholders.
Why the Old Playbook Won’t Work
Here’s the thing about traditional credit assessment: Banks still lean on bureau scores – it’s like trying to predict tomorrow’s weather by looking at last month’s newspaper.
This approach doesn’t just miss problems; it’s like using a sledgehammer to hang a picture frame. Blunt provisioning buffers treat all risk the same way, which means you’re either choking good business or ignoring real dangers. What’s needed isn’t bigger hammers; it’s the kind of precision that can tell whether Raju’s electronics shop is having a temporary cash crunch or heading for permanent trouble.
The Tech Shift: From Buffers to Brains
This is where AI, alternative data, and India’s digital infrastructure, Account Aggregator (AA) and Unified Lending Interface (ULI), flip the entire script. It’s like upgrading from a flip phone to a smartphone, except the smartphone can predict the future:
- Cash flow becomes crystal clear: AI models can read GST returns, AA-sourced bank flows, and UPI payment histories like a fortune teller reading palms, except with actual data. When a borrower’s cash flows start getting choppy, you get 60-day early warnings instead of 90-day heart attacks.
- Early intervention beats late reaction: Instead of waiting for someone to miss their EMI, like waiting for a delayed train, machine learning flags anomalies in spending patterns within days. It’s like having a financial fitness tracker that alerts you before you pull a muscle, not after you’re already limping.
- Micro-segmentation at massive scale: ULI and AA let lenders create risk strategies as precise as a master chef’s recipe. You can keep profitable customers happy while filtering out trouble, instead of using the same approach for everyone.
- Collections that actually work: AI can rank delinquent accounts by recovery probability and suggest exact treatment paths. No more playing pin the tail on the donkey with collection strategies, hoping something works.
RBI’s Balancing Act
The regulator has been playing the role of that wise parent who tells teenagers they can stay out late, but they’d better act responsibly. Rolling back some risk-weight add-ons in February 2025⁷ wasn’t a random policy decision; it was a clear signal: “We want credit to grow, but we also want you to be intelligent about it.”
Tools like AA and ULI provide exactly the kind of transparent, explainable infrastructure that makes regulators smile. You can show your homework, explain your reasoning, and prove you’re managing risk instead of just crossing your fingers and hoping for the best.
The Bottom Line
Indian banking is growing. The sector is moving from the wild growth phase to the mature management phase, where sustainable returns matter more than eye-popping growth rates. Private banks face margin compression and rising provisions⁸, but they’re also building more resilient business models.
The infrastructure is already here, humming along like a well-oiled machine. The data flows are real, Account Aggregator is scaling faster than viral Reel videos, and AI can process thousands of risk signals before you finish reading this sentence.
The playbook sounds straightforward enough: cash-flow aware underwriting, integrated fraud screening, early-warning systems, and collections that learn from their mistakes. But executing this requires the kind of technology infrastructure that can handle real-time data flows, explain its decisions to regulators, and adapt quickly when market conditions shift.
This is exactly what modern decisioning platforms must deliver: cash-flow aware underwriting, integrated fraud and credit decision engines, and early-warning systems that actually work in the real world, all with the explainability and compliance features regulators increasingly demand.
The goal remains simple: keep growth responsible, keep credit costs contained, and keep good customers in the system. It’s just the execution that gets complicated. But then again, if banking were easy, everyone would be doing it well.
Because in the end, the race won’t be won by whoever can absorb the most losses, it will be won by whoever prevents them in the first place. And honestly, that’s a much more interesting game to watch.