Bajaj Finance's AI listened to 2 crore customer calls last quarter. It converted 5.2 lakh of them into structured credit data. Then it generated 1 lakh new loan offers for customers the company previously knew nothing about.
That is not a pilot. That is production.
The Q3 FY26 earnings call, reported by Outlook Business in February, included a disclosure from MD Rajeev Jain that most credit heads at mid-size NBFCs probably read past without stopping: "Loan disbursements through AI-powered call centres stood at about ₹1,600 crore" — roughly 10% of the ₹16,545 crore disbursed that quarter. And Jain added they're planning to scale to 100 million calls next year.
India's NBFC sector is growing at 17% annually. That headline number hides a split.
At the top, Bajaj Finance, Tata Capital, and Poonawalla Fincorp are running enterprise-wide AI — underwriting co-pilots, voice analytics in collections, agentic systems across the lending lifecycle. Tata Capital's Q3 FY26 commentary talked about "moving from AI pilots to enterprise-wide deployment" across credit, sales, and operations.
At the middle and lower tiers, the picture is different. Credit teams still pull bureau reports manually. CAM notes take 3–5 days to prepare. Loan turnaround for an MSME borrower is 14–30 days.
An EY report from early 2026 found that 42% of Indian financial institutions are actively investing in AI, with 74% running proofs-of-concept. But investing in a POC and deploying ₹1,600 crore in loans through an AI-powered channel are different things. One is a budget line. The other is a business model.
Ashutosh Taparia of CredAble, who works with lenders on AI deployment, put a number on the gap: "What was once a 14-day manual process can become a 14-minute autonomous workflow, with human-in-the-loop oversight." He was talking specifically about MSME lending — the segment where documentation is messiest and human judgment is most stretched.
The efficiency claim isn't theoretical. His teams reported "over 40% efficiency gains and close to 90% AI-led accuracy" in their deployments.
The counterintuitive point here: the lenders most resistant to AI underwriting are often the ones with the highest credit costs. The argument against AI adoption is usually about accuracy risk. The data suggests the actual risk is the opposite — manual processes are slower, more biased, and less consistent than well-governed AI systems.
RBI 2026 guidelines now require documented bias testing across protected characteristics before any AI underwriting model enters production. That's a sensible guardrail. But it is a guardrail on a road most NBFCs haven't yet turned onto.
Here is the practical problem for a ₹500–2,000 crore AUM NBFC operating in MSME or personal lending.
Bajaj Finance has 46 million face-match validations running. It knows its returning customers faster than a credit officer can pull up a file. Its AI is not just cutting costs — it is identifying new borrower segments that manual processes miss entirely. Those 1 lakh new loan offers didn't come from a marketing campaign. They came from the data in calls that humans couldn't process.
When Bajaj Finance or Tata Capital moves into a segment you thought was yours — gold loans, vehicle finance, MSME working capital — they arrive with better data, faster decisions, and lower credit costs. The competitive moat is not just capital. It is information processing at scale.
A credit head at a mid-tier NBFC once told me their biggest risk wasn't NPA — it was adverse selection. The well-documented borrowers go to the lenders who can process them fastest. They get left with the borrowers no one else wanted.
AI underwriting doesn't just lower costs. It determines which borrowers you even see.
What is AI credit underwriting and how are Indian NBFCs using it? AI credit underwriting uses machine learning to score borrowers, process documents, and generate credit decisions with minimal human intervention. In India, large NBFCs like Bajaj Finance and Tata Capital have moved from pilots to enterprise deployment, using AI to analyze call data, match identity documents, and disburse loans faster than traditional processes allow.
How much of Bajaj Finance's lending is AI-driven? In Q3 FY26, Bajaj Finance disbursed ₹1,600 crore through AI-powered call centres — approximately 10% of its total ₹16,545 crore disbursed that quarter. The company's AI also analyzed 2 crore customer calls and created 1 lakh new credit offers from previously unstructured data.
Does RBI allow AI-based credit scoring for NBFC lending? Yes. RBI's 2026 guidelines permit AI-based credit underwriting but require documented bias testing across protected characteristics before any AI model enters production. The regulatory expectation is that AI systems are explainable, auditable, and governed — not a black box.
Abhishek Gupta is Co-Founder at Dekrypt Labs, building BIOS — a Business Intelligence Operating System for Indian businesses. dekryptlabs.com