AI can speed business loans, but underwriters stay in charge: HyperVerge's Kedar Kulkarni

AI-based identity-verification and fraud-prevention company HyperVerge is betting that artificial intelligence (AI) can reduce hours of manual work in business-loan underwriting into minutes, while leaving final credit decisions with people, cofounder and chief executive Kedar Kulkarni told ET AI.

The company on September 9 launched three underwriting agents at the Global Fintech Fest 2026 in Mumbai. The first pulls information from bank statements, GST filings and tax returns, the second assists video-based personal discussions and the third screens businesses and their promoters against corporate filings, litigation records and sanctions lists, according to the company.

More than half of business-loan applications reach lenders through direct selling agents (DSAs), who often send a file to two or three lenders, Kulkarni said. Since lenders may require different documents, exchanges among the DSA, salesperson and credit manager can consume a full day before underwriting begins, he added.


Also Read: If attackers are using AI, companies will have to use AI to defend themselves: Razorpay CEO

The credit manager then reconstructs the business’s finances, checks whether it fits a lending programme and may arrange a video call or site visit, Kulkarni said. The lender also examines the business and its directors against court records, police cases and other sources, he added.

A background check can take about three hours because a manager may need to search multiple services and distinguish the applicant from people with the same name, Kulkarni said.

“A process that now takes three hours can be presented on a single screen with a green, amber or red assessment. The screen shows all 15 checks, which ones have passed and which ones require review,” he said.

HyperVerge tested the suite with about 10 mid-sized lenders, while three have put its video discussion agent into production, the company said in a statement. It said financial assessments fell from two or three hours to five minutes in pilots, while credit appraisal memos that took four hours were created in less than 10 minutes.

Kulkarni said the tools are also going live with lenders including Cholamandalam Investment and Finance Company and L&T Finance. Business loans currently take between two and seven days and HyperVerge is working with lenders to bring some decisions down to the same day or reduce seven-day cases to two days, he said. The technology is also live in the US, where HyperVerge has seen credit costs fall by more than 10%, he added.

Despite the time savings, Kulkarni ruled out allowing an agent to approve or reject a complex business loan by itself.

“We are not looking at removing human involvement at all. These are complex loans involving too many data points, so a human must remain involved,” he said.

An underwriter’s note that takes two hours to prepare could be generated by AI in about a minute and reviewed in five minutes before the existing approval process continues, Kulkarni said. He described the system as a “cricket match’s highlight reel” that directs the underwriter to important information instead of creating new facts.

To contain hallucinations, HyperVerge uses output guardrails, links flagged issues to the underlying document or record and retains an audit trail for regulatory review, Kulkarni said. Lenders normally start with a small group of underwriters who compare AI results with their old process before expanding deployment, he said. They also monitor the repayment behaviour of borrowers handled through the workflow, he added.

Also Read: GFF 2026: BharatPe launches agentic AI assistant for merchants, connects to over 60 live systems

The system sits on top of lenders’ existing infrastructure rather than replacing it, Kulkarni said. HyperVerge made it modular because lenders want to test one task, such as background checks, video intelligence or document review, before widening its use, he added.

The company cited estimates that only 41% of India’s 8.7 crore registered micro, small and medium enterprises have accessed formal credit and put the MSME credit gap at roughly Rs 25 lakh crore. Kulkarni said lower underwriting costs could make smaller loans viable, while AI-assisted video checks could help lenders serve areas where they lack branches.

“If underwriting costs Rs 25,000, a lender may not be able to offer a Rs 5 lakh loan. As the cost of underwriting falls, more people naturally gain access to credit,” he said.

Kulkarni expects genuinely pre-approved products to become more automated when lenders already hold extensive financial data on borrowers. For complex and asset-backed credit, however, he expects human review to remain.

“I believe financial-services systems will always have a human over the loop because regulators will require it,” he said.