AI-led underwriting is the missing link in India’s MSME credit story

India has spent a decade building the finest financial plumbing in the world. The Unified Payments Interface (UPI) clears well over 20 billion transactions a month, close to half of all real-time payments on the planet. Aadhaar gave a billion people an identity. The Account Aggregator network now lets small businesses share verified bank and GST (goods and services tax) data in seconds, with consent, no PDFs, and no branch visits.

And yet the one number that matters most has barely moved. India’s MSME (micro, small, and medium enterprise) credit gap sits at Rs 25-30 lakh crore. We have 87 million registered MSMEs, and only 36 million have ever taken a formal loan. The share of first-time borrowers in fresh lending has actually fallen—from 52% three years ago to 42% today. We may have built the rails, but we are not moving enough people down them.

So, where is the blockage?


It is not payments. It is not identity. It is not even data anymore. The blockage sits at the very end of the pipe; in the one place we have left untouched. It is underwriting.

The old model was built for someone else

Traditional underwriting was designed for large, formal borrowers. It asks for three years of audited financials, a property to pledge, and a long credit history. A mid-sized company can produce all three. The kirana store owner, the fabricator with two machines, and the woman running a catering business from her kitchen cannot.

A bureau score makes this worse, not better. It tells you how someone handled debt in the past. It says nothing about whether the business in front of you can afford a new loan today. For a first-time borrower with a thin file, the score is often blank. So, the application is declined; not because the business is weak, but because the model cannot see it.

This is the quiet tragedy of MSME lending in India. Millions of viable businesses are invisible to the very system built to fund them.

What AI-led underwriting actually changes

The shift is simple to state. AI-led underwriting stops asking, “what can you pledge?” and starts asking “What does your money actually do?”

Pull 12 months of bank transactions through the Account Aggregator. Read the GST invoices. Look at the UPI inflows, the rhythm of receipts, the seasonality, and the way past dues were cleared. A model can turn all of that into a real picture of a business in minutes, for a loan of Rs 50,000, at a cost that finally makes small-ticket lending viable.

That last point is the one that most lenders miss. Banks avoid micro-loans not only because of risk but also because of cost. When it takes a week of manual document collection to assess a one lakh loan, the economics never work. AI collapses that week into minutes. Suddenly the small borrower is worth serving.

This is where the human comes back in

I do this for a living, so let me be clear about what AI-led underwriting is not. It is not a machine replacing your credit officer.

The machine does what machines do best. It reads the data, spots the patterns, flags fraud, and holds the line on consistency. That frees your best people to do what only people can. To look at an unusual cash flow and understand the story behind it. To weigh the local context a model will never capture. To approve a deserving borrower the score would have rejected, and to explain that decision to a regulator with a straight face.

The strongest lending calls in this market will not come from an algorithm alone or from a human working blind. They will come from a human with a machine doing the heavy lifting behind them. That is how you scale judgment instead of trading it away.

One caution

Cheap and fast is not the same as good. Early stress is already showing in some unsecured MSME pockets. The answer is not to slow down. It is to underwrite on real cash flow rather than chase volume on thin signals, and to make sure every automated decision can be explained in plain language to the borrower and the regulator alike.

The rails are built. The data flows freely. The one piece still stuck in the past is the decision at the end of the line. Fix underwriting, and the Rs 25 lakh crore gap stops being a statistic and starts becoming the largest untapped market in Indian finance.

That is the missing link. And it is well within reach.

Joydip Gupta is Head of APAC at Scienaptic AI, Chandan Pal is Chief Marketing Officer at Scienaptic AI. Views are personal