We apply AI across the lending lifecycle, and underwriting is a core focus: Markandey Upadhyay, Piramal Finance

As artificial intelligence reshapes the financial services landscape, the real differentiator for lenders is moving beyond technology adoption to the ability to apply intelligence meaningfully across the lending lifecycle. For Piramal Finance, this means moving towards an AI-native operating model, where predictive analytics, machine learning and agentic AI work together to strengthen credit decisions, improve customer experience and enhance operational efficiency. With a focus on understanding diverse borrower profiles at scale, the company is reimagining underwriting, fraud management and collections through proprietary AI capabilities. In an exclusive conversation with Express Computer, Markandey Upadhyay, Chief Digital & Analytics Officer, Piramal Finance, discusses the company’s AI-led transformation, its approach to building intelligent lending systems and the next phase of its technology roadmap.

For decades, success in lending has been built on capital, distribution and risk management. Is AI emerging as the fourth pillar, and how is Piramal Finance leveraging it across the lending lifecycle?

Capital, distribution and risk management will always remain fundamental to lending, but AI is increasingly becoming a force multiplier across all three. At its core, lending is a business of understanding. Every borrower has a unique financial journey, aspirations and circumstances. AI enables us to derive deeper insights, identify patterns faster and support better decision-making at scale.

At Piramal Finance, we apply AI across the lending lifecycle, with underwriting being one of our most important areas of focus. We have been using proprietary machine learning models for several years to assess the probability of default and strengthen credit decisions. Over time, we have developed multiple models that evaluate different aspects of a borrower’s profile, including credit bureau records, banking transactions, customer demographics and asset-related information in the case of secured lending. We also use AI models to identify identity and document fraud. We have progressively moved from individual models to multiple models, followed by meta-models, and are now orchestrating them through an agentic approach that we call Agentic-Based Credit Underwriting (ABCD). This brings together different aspects of AI to enable a more holistic credit assessment.

For instance, bank statements have become increasingly complex with the proliferation of UPI transactions. Analysing these statements manually is challenging, particularly when it comes to identifying anomalies or detecting tampering. AI can examine large volumes of transaction data and identify inconsistencies that may not be visible to the human eye. The real value of AI is not simply its ability to process more information, but its ability to help lenders understand customers better and make more informed decisions.

What differentiates an AI-enabled organisation from an AI-native one? How is Piramal Finance building towards an AI-native operating model?

The difference is similar to the distinction between digitising a process and reimagining it altogether. An AI-enabled organisation uses AI to improve existing workflows, whereas an AI-native organisation designs its workflows, decisions and customer journeys with intelligence embedded from the outset. At Piramal Finance, we have a few structural advantages. When we significantly expanded our retail business around five years ago, we consciously chose to build a cloud-native organisation. Our infrastructure is cloud-based, and we own the source code for our core lending systems, including our loan origination system (LOS), loan management system (LMS) and collections platform.

This gives us the flexibility to integrate multiple AI models at different stages and redesign our lending journeys as technology evolves. With third-party platforms, integrating customised models and changing workflows can be considerably more challenging. Another important factor is the adoption of AI across the business. Our CEO, Jairam Sridharan, started his career as a data scientist and analyst, which has helped establish a strong data-driven culture within the organisation. Business teams actively explore how data science and AI can address operational challenges. This also makes adoption easier because the demand for AI solutions comes from business users themselves. Talent is the third critical component. We have invested in building a strong team of AI scientists and data science professionals to develop sophisticated, business-aligned solutions.

Our approach is to work on real business problems rather than limiting ourselves to proof-of-concept projects. Models may not deliver the desired results in the first iteration, but we continuously refine them based on feedback from business users and take them towards implementation at scale.

India’s lending landscape is diverse, with customers across geographies and income segments having different financial needs. How can AI help financial institutions understand and serve these customers better?

This is where the opportunity becomes particularly exciting. A borrower in Jaipur, Coimbatore, Indore or Mumbai may all be seeking credit, but their financial circumstances, aspirations and requirements can be very different. Traditional approaches often require institutions to standardise processes to achieve scale. AI changes that equation. For a nation as diverse as India, AI’s greatest promise lies in understanding millions of unique stories at scale. It allows institutions to move closer to treating every customer as an individual, even while serving millions. That is what building AI for Bharat means in practical terms. It is not about creating models for the sake of technology, but about building systems that understand local realities, support informed decision-making and make financial services more responsive to individual needs. Predictive analytics also allows us to anticipate customer requirements by identifying patterns in their financial behaviour. For instance, banking transactions can reveal whether a customer is preparing for higher education abroad through payments towards entrance examinations, university applications or visa-related expenses. By analysing these indicators collectively, a financial institution can anticipate a potential requirement for an overseas education loan.

We also use predictive analytics for risk assessment. If a customer’s banking transactions indicate significant activity across gaming or gambling platforms, our models can identify the associated risk. This allows us to investigate further or exercise caution before extending credit. We have implemented such a use case at Piramal Finance. The future of lending will belong to institutions that can combine the efficiency of scale with the relevance of personalised understanding.

How is Piramal Finance translating its AI investments into measurable business outcomes, particularly in underwriting, fraud detection and collections?

As of Q2 FY2027, Piramal Finance serves more than 6.5 million customers through 838 branches across 621 cities and manages over ₹1.13 lakh crore in Assets Under Management.  

Through Piramal.ai, intelligence is increasingly embedded across customer journeys and operational processes. AI-assisted underwriting has analysed nearly 1.94 lakh bank statements, while document fraud intelligence has screened more than 6 lakh documents. AI-driven fraud monitoring has generated alerts across over 20 lakh cases, strengthening risk and control frameworks. Voice AI enabled ₹102 crore in cross-sell disbursements, while ARYA, the AI-enabled work assistant, resolved ~5.6 lakh queries, up 45% yoy. AI-powered collections are contributing approximately ₹754 crore in monthly hands-free collections, while digital customer resolutions have reached 74%.  

Piramal Finance continues to scale GenAI adoption through Piramal.ai, embedding AI across core business operations and moving beyond pilots to day-to-day integration. This is enhancing productivity, customer experience and decision-making across the organisation. 

What is encouraging is that these initiatives are no longer being viewed as pilots or experiments. They are becoming part of how the organisation operates. As intelligence gets embedded across underwriting, servicing, fraud management and collections, it creates a foundation for a more agile, responsive and future-ready institution. Overall, Piramal.ai is delivering measurable gains across productivity, business growth, risk management, collections, customer and stakeholder engagement.

What are the key technology priorities on Piramal Finance’s roadmap over the next few months?

Our biggest priority is transforming underwriting. We are moving from single credit-risk models to multiple risk models, then meta-models and now towards the next generation of agentic-based underwriting. This is a significant transformation that involves not just developing models but also building the right ecosystem, teams and processes to support the new underwriting journey. The second priority is transforming our fraud management architecture. Currently, fraud detection can be fragmented across different functions, including document verification, customer profiling and investigation processes.

We are bringing customer information together on a single platform and integrating an AI engine to assess risk. Based on the assessment, the system can determine whether a customer can proceed with the lending journey, should be rejected or requires enhanced due diligence. This will help us create a more integrated and intelligent fraud management framework. Voice AI is our third major priority. Rather than relying entirely on third-party solutions and off-the-shelf models, we are building our own capabilities by orchestrating speech-to-text, text-to-speech and AI-based interaction systems using our proprietary data. This initiative is at an early stage, but we aim to scale it significantly.

Alongside these priorities, we are also focusing on what we call ‘Everyday AI’. We want employees across the organisation to use AI to improve their daily productivity. We have introduced tools that allow employees to build their own dashboards, technology tools and AI agents. Even non-technical business users have created hundreds of such tools to address their everyday requirements. Our objective is to make AI accessible across the organisation, enabling employees to become more productive while ensuring that enterprise data privacy and security requirements are addressed.

Looking ahead, what will distinguish the leaders in the next decade of financial services?

The winners will not necessarily be those with access to the most advanced AI models. Technology advantages tend to narrow over time. What will truly matter is an institution’s ability to combine intelligence with trust, scale with relevance, and automation with human judgement. Financial services has always been a people business, and that will not change. What will change is the ability of institutions to understand customers more deeply and respond more intelligently. Organisations that use AI not just to automate processes but to create a better understanding of the people they serve will be better positioned to differentiate themselves. In a country like India, that is perhaps the most important opportunity of all.

Comments (0)
Add Comment