India’s NBFC sector is undergoing a structural transformation. Lending is steadily shifting from relationship-driven decision-making towards intelligence-led models powered by data, automation, and embedded finance. At the same time, regulatory pushes around financial inclusion, co-lending, and governance are reshaping how lenders expand access to credit while managing risk.
For lenders operating across MSME finance, affordable housing, gold loans, and construction finance, the challenge is no longer simply scale—it is how to scale responsibly, faster, and with sharper credit intelligence.
At Capri Global Capital, this shift is increasingly being driven through AI-enabled underwriting, predictive collections, real-time risk monitoring, and cloud-native infrastructure under the leadership of Tarun Aggarwal, Group CTO, Capri Global Capital. Rather than treating AI as a standalone innovation layer, the company is embedding intelligence across lending, customer engagement, engineering, fraud detection, and operational resilience.
According to Aggarwal, the broader industry itself is moving toward a new operating model.
“The biggest shift is that lending is moving from relationship-based decision-making to data-driven credit intelligence,” he says.
Co-lending and the democratisation of credit
One of the biggest forces reshaping NBFC growth today is the co-lending ecosystem encouraged by the regulator.
Traditionally, NBFCs have been highly effective at reaching underserved customer segments—small business owners, first-time borrowers, affordable housing seekers, and customers outside the formal banking ecosystem. However, their cost of funds tends to be higher than that of banks.
The co-lending model attempts to solve this structural challenge. Under the framework, banks can fund up to 80% of a loan while NBFCs contribute the remaining share, helping customers access credit at lower borrowing costs while enabling NBFCs to expand responsibly.
Aggarwal sees this as particularly important for “new-to-credit” borrowers. “Inclusion is not possible unless millions of customers outside the traditional credit ecosystem are brought into formal finance,” he says.
For Capri, this becomes particularly relevant across MSME lending, affordable housing, and gold loans, where customer acquisition and contextual underwriting matter deeply.
Speed, trust, and intelligence in gold lending
In segments such as gold loans, operational speed becomes a competitive differentiator. “If a customer walks into a gold loan branch, speed is everything,” Aggarwal says.
According to the company, gold loan processing—from customer entry to loan disbursal—is completed in under 30 minutes. The process includes gold valuation, fraud checks, authentication, and digital documentation.
But behind that speed sits a broader technology stack. Across its branch ecosystem, the company uses centralised monitoring and AI-assisted controls for vault access, fraud validation, and operational visibility. Branch operations are also designed to remain largely paperless, with digital agreements and transparent loan documentation replacing traditional manual processes.
The objective, Aggarwal argues, is not simply faster lending, but greater trust and repeat customer engagement.
AI is moving from experimentation to operational reality
While much of the financial services industry continues experimenting with AI, Aggarwal believes the conversation inside regulated lending is increasingly shifting toward measurable operational outcomes.
“At Capri, AI has moved from a buzzword to an operational reality,” he says. The company has invested in an internal AI infrastructure, supported by air-gapped computing environments designed to align with security and compliance expectations in a regulated sector. A dedicated AI and data science team of roughly 25 members is currently expanding further.
Rather than replacing human judgment, Capri positions AI as an intelligence layer supporting decision-making. In underwriting, machine learning models analyse behavioural and data signals to help assess creditworthiness—particularly for customers with limited formal credit history.
For “new-to-credit” borrowers, this becomes especially valuable. Instead of relying exclusively on conventional credit scores, systems surface hidden risk indicators and contextual insights that are then reviewed by credit teams before final approval.
“The final decision always remains with humans. AI surfaces hidden patterns that might otherwise be missed,” Aggarwal explains.
A similar approach is visible in collections. The company deploys early warning models, probability-to-pay frameworks, and communication preference engines to predict potential repayment stress and personalise engagement strategies. Some borrowers may respond better to calls, others to SMS or WhatsApp-based reminders. This allows collections teams to move from reactive recovery models toward more predictive engagement.
Voice intelligence, fraud monitoring, and the rise of contextual AI
One of Capri’s more ambitious focus areas lies in voice intelligence and conversational AI.
The company has built systems capable of transcribing customer interactions, analysing contextual signals, and identifying potential fraud indicators in near real time. Customer conversations are converted into text and assessed to improve collections quality, audit readiness, and compliance monitoring.
The company is also piloting conversational bots for collections workflows. These systems can handle routine customer interactions autonomously while escalating seamlessly to human agents whenever contextual complexity increases.
Aggarwal believes this “human-in-the-loop” approach will become increasingly important as AI adoption matures across financial services.
The future, however, may lie beyond today’s large language models.
Capri is investing in specialized small language models (SLMs)—compact, domain-specific AI systems designed for faster, localised, and potentially offline decision-making.
Unlike large external AI systems, these models can run closer to the edge, reducing latency, lowering infrastructure costs, and keeping sensitive data within controlled environments. For regulated sectors like lending, Aggarwal believes this could eventually become an important trust and compliance differentiator.
Building compliance-first AI in a regulated industry
For NBFCs, innovation cannot exist separately from compliance.
Aggarwal repeatedly frames technology decisions through what he describes as a “compliance-first” architecture aligned closely with regulatory expectations.
At Capri, technology deployments are mapped in line with regulatory frameworks, supported by explainability principles, automated quality checks, and DevSecOps pipelines that validate code before production release. The company also operates a 24×7 security operations capability and is preparing for evolving data privacy requirements under India’s digital privacy regulations.
“Compliance and security are not afterthoughts—they are embedded by design,” he says. This becomes particularly important as AI adoption expands.
Aggarwal believes fraud risks, synthetic identity attacks, and manipulated KYC workflows will become more sophisticated as generative AI capabilities improve. The implication for lenders: cybersecurity must shift from reactive defence to proactive anticipation.
Engineering for resilience and scale
Beyond AI, Capri’s technology modernisation story is equally centred on resilience.
Over the last several years, the company has moved core engineering capabilities in-house, building its own loan origination systems while strengthening cloud-native, multi-region infrastructure and operational redundancy. An engineering organisation of more than 180 professionals now supports platform development internally.
Operational excellence programs monitor customer-facing performance metrics continuously, helping teams identify latency bottlenecks, reduce outages, and improve digital responsiveness.
The company has also invested significantly in business continuity and disaster recovery preparedness. According to Aggarwal, recovery time objectives have improved by more than 80%, while recovery point objectives are now close to near real-time through multi-cloud resilience frameworks.
At the same time, automated engineering pipelines supported by a QA team of roughly 20 specialists increasingly detect software defects earlier in development cycles, reducing operational risk and improving release quality.
The future of lending will be intelligent—but human-led
Looking ahead, Aggarwal sees agentic AI, edge intelligence, and voice-first automation becoming increasingly influential across financial services.
Yet he remains clear on one principle: humans will remain the ultimate decision-makers.
“The company that builds the right guardrails today will be the one that scales safely tomorrow,” he says.
For Capri, the long-term ambition is closely tied to financial inclusion itself. Technology, in this vision, is not simply about automation or efficiency. It is about making formal credit more accessible to millions of Indians while balancing compliance, trust, and responsible growth.
As India’s lending ecosystem becomes increasingly digital, the competitive advantage may no longer belong only to institutions with deeper balance sheets—but to those capable of combining intelligence, speed, resilience, and regulatory discipline at scale.
