How AI Helps Piramal Finance Grow Without Opex Increase

Piramal Finance’s Chief Data and Analytics Officer, Markandey Upadhyay, says by embedding AI, ML and decision science models into underwriting, the company has been able to keep credit costs firmly under control.

Non Banking Finance Company, Piramal Finance (PFL), has widely adopted AI and GenAI across the processes in the lending journey for many of its products. The NBFC has been able to reap comprehensive benefits from its investments in AI and GenAI. 

“Our successful deployment of traditional and generative AI across the businesses reflects our commitment to operational excellence and technology-led scale. Our improved retail opex-to-AUM ratio—better than initial guidance—has been driven by AI investments now embedded across risk, productivity, controls, and more,” says Ajay G. Piramal, Chairman, in his message to the shareholders in the annual filing of the company for FY24-25.

AI is separating the potentially risky borrowers

PFL’s AI risk models employ custom product-specific scorecards instead of standard one-size-fits-all models, which dilute performance. Product-specific scorecards detect unique behaviours, better predicting default risk, informed the company in its FY25 filing.

The AI based tool Piramal Finance uses looks at different groups of borrowers and predicts how much debt they can handle and still pay back on time. When Piramal tested the tool, it showed that some borrower groups were 3 to 6 times riskier than others. This means the tool is good at separating safe borrowers from risky ones, so Piramal can avoid lending to people who are very likely to default, says the exchange filing.

Piramal Finance’s Chief Data and Analytics Officer, Markandey Upadhyay, speaking with FE CIO echoed his Chairman’s statement, saying, 'We have been able to grow at a consistent speed without a similar increase in opex and manpower by making AI central to our way of working.' Excerpts below.

The Annual Report for FY24-25 mentions strong investments by the company to become an Al-first, data-led organisation. Specifically, Al, ML, and decision science are enhancing customer experience, detecting fraud, and strengthening lending— resulting in improved retail Opex-to-AUM ratio. Please elaborate. 

At Piramal Finance, we see AI/ML, decision science and agentic AI solutions as an integral part of our way of doing business. We have built models that touch every stage of the lending journey. 

We enhance productivity of our sales staff with automated document verification at the login stage. We underwrite better using our proprietary risk and fraud detection tools, driving disbursement and growth, along with better customer experience. We are digitising bank statements and other documents to cut processing times and surface deeper insights, while anomaly detection strengthens fraud prevention. We are working extensively for our people and have created our internal AI assistant ARYA, which is transforming how our staff work, giving them live visibility into disbursements, incentives, and even promotion eligibility. These investments help scale disbursements while keeping operating expenses under control.

The Al-based credit underwriting model is noted to be patent pending. What is the key differentiator that makes this model innovative and patent-worthy?

With rising unsecured delinquencies, we were seeing that a segment of customers were defaulting due to over leverage. This leverage is accrued after taking our loan, even though at the time of underwriting, they were not over-leveraged. Our model aims to identify such customers, who have a high future over-leverage risk. This model is built on our proprietary data and is an industry first. This capability allows us to screen out the segment of customers who get over leveraged and end up having 3x-6x the risk. These models provide us an added benefit over and above our traditional AI risk models. While, industry is managing current leverage risk, this model is a step forward, in pre-empting future leverage, and hence we have filed a patent for this.

How would you sum up the overall benefits derived from Al and GenAl across FY24-25?

We have been able to grow at a consistent speed without a similar increase in opex and manpower by making AI central to our way of working. By embedding AI, ML and decision science models into our underwriting, we’ve kept credit costs firmly under control. What we see here is just the beginning, and as we scale our deployments and with the evolution of AI from LLMs to Agents, we expect greater benefit in the future.

Empower your business. Get practical tips, market insights, and growth strategies delivered to your inbox

Subscribe Our Weekly Newsletter!

By continuing you agree to our Privacy Policy & Terms & Conditions