New Delhi, Jan. 23 -- From agentic underwriting to patent-pending risk models, Piramal Finance, a wholly owned subsidiary of Piramal Enterprises Limited, and a housing finance company, is embedding artificial intelligence deep into the core of its lending operations. In an interview with TechCircle, Saurabh Mittal, Chief Technology Officer (CTO), Piramal Finance, explains how the NBFC is using artificial intelligence (AI), machine learning and data science to drive faster decisions, stronger risk discipline and scalable growth-while operating within the tight guardrails of financial services regulation. Edited excerpts.
How is Piramal Finance using digital and AI-led technologies to transform lending and risk management across retail and wholesale businesses?
At Piramal Finance, digital transformation is about embedding intelligence deep into the lending lifecycle rather than layering technology on top. Over the last few years, we have built strong AI, machine learning and decision science capabilities that span onboarding, underwriting, disbursement, monitoring and collections. At the front end, AI-led agentic systems triangulate information across multiple data sources-KYC records, bank statements, salary slips and application data-dramatically reducing manual file processing. This has shortened turnaround times and lowered operational overhead for both customers and internal teams.
In underwriting and risk, we deploy multiple supervised models across credit assessment, fraud detection, income estimation and policy evaluation. Classical machine learning models work alongside AI agents to provide underwriters with richer, more reliable signals at the point of decisioning. AI-driven document synthesis further helps teams interpret complex profiles and large datasets. Together, this integrated, AI-native approach enables scalable growth with strong risk discipline-without a commensurate increase in manpower or operating costs.
What role do AI, machine learning and data analytics play in core operations today?
These are foundational capabilities for us. Across teams and business lines, employees interact with AI-driven systems daily. We use machine learning to differentiate risk and optimise collections, agentic AI to validate documents, advanced models for fraud detection, and computer vision for customer identification and verification. Internally, teams rely heavily on our AI assistant, ARYA, which supports day-to-day workflows-from tracking incentives and sales performance to checking lead status, planning work and referencing internal policies. These workflows are increasingly AI-driven and embedded into how we operate. All of this sits on a modern, resilient technology stack that supports scale and low latency. Business users, engineers and AI practitioners work as one unit to solve real problems. While most AI projects globally fail, our belief system, operating model and ecosystem help us stay in the small minority that delivers impact.
Can you share examples where technology has delivered measurable business impact?
One clear example is our AI-driven underwriting framework. Instead of one-size-fits-all models, we deploy a bouquet of product-specific scorecards covering credit, fraud, income, eligibility and asset quality. This has significantly improved risk separation and enabled more confident credit decisions. Another innovation is our patent-pending leverage risk model, which identifies customers who may become over-leveraged post-disbursement-even if they appear healthy at underwriting. Customers flagged by this model show materially higher default risk, making it a powerful enhancement to traditional checks. We have also introduced agentic systems that extract and triangulate information from documents-tasks that were previously manual, time-consuming and error-prone. AI allows teams to move faster and focus their expertise on complex work. These outcomes are possible because AI systems are tightly integrated into the decision flow, not deployed as standalone tools.
How are you leveraging generative AI in a regulated financial services environment?
We are very deliberate. At Piramal Finance, GenAI is used to augment human intelligence, not replace it. Our primary focus areas are internal productivity, insight synthesis, workflow simplification and developer enablement, all within strong governance guardrails. Most current use cases are inward-facing. We have built robust oversight, validation and monitoring frameworks to identify and mitigate risks proactively. AI-generated outputs are rigorously reviewed before production deployment and continuously monitored thereafter to ensure compliance and alignment with our risk appetite. This ecosystem allows us to harness GenAI responsibly in a regulated environment.
How do you balance innovation with cybersecurity, compliance and data privacy?
For us, innovation and trust go hand in hand. We have invested deeply in governance and end-to-end model lifecycle management. Dedicated annotation and hindsight teams track model performance, drift and explainability over time. In fraud prevention, for instance, we use in-house computer vision models to detect document tampering-spotting anomalies that are difficult for humans to identify consistently. Importantly, AI alerts always go through human review. Cybersecurity, compliance and data privacy are non-negotiable, and the same rigor applies to AI systems as to any other critical infrastructure.
What are your top technology priorities for the next 2-3 years?
First, we want AI to become an express lane for business execution-driving customer acquisition, credit sanctions and conversions, while acting as a full-time assistant for frontline teams. Second, we aim to extend AI-led productivity gains beyond engineers to corporate and support functions, eliminating repetitive tasks and freeing teams for higher-value work. Third, we are embedding AI at the core of decisioning, especially underwriting. AI can synthesise disparate data sources transparently, improving outcomes. For simpler cases, AI will manage end-to-end workflows, allowing human experts to focus on complex scenarios. In parallel, we will continue strengthening data, governance and monitoring foundations to support scale and regulatory confidence.
How are you building a tech-first, AI-native culture and tapping India's talent pool?
Our key learning is that AI adoption succeeds when driven by users. Many initiatives originate from business teams that see value and seek enhancement, creating strong bottom-up momentum. We operate with multidisciplinary teams combining engineering, agentic AI, small language models and domain expertise. We also run training programmes that enable even non-coders to build solutions on internal platforms. Being AI-native is not about a single model-it's about an ecosystem where business, technology and data collaborate from day one.
How are partnerships important to your innovation strategy?
Yes. We work with hyperscalers like AWS, Azure and Google Cloud, and with specialist partners in areas such as voice-based AI. But partnerships alone are not enough. The real value comes from combining external expertise with strong internal ownership to co-create and operationalise solutions that are truly fit for our customers and business.
Published by HT Digital Content Services with permission from TechCircle.