How India’s DPI Stack is Building the Foundation for AI in BFSI

India built the infrastructure rails before the AI race began, and that sequencing might end up being the biggest changemaker.

Trust built BFSI. Now it’s slowing it down. The BFSI sector stands to tremendously benefit from the adoption of AI, yet it remains one of the hardest sectors to transform. The very characteristics of the industry that make it trustworthy- regulatory oversight, data sensitivity, relationship-driven operations, are proving to be the biggest barriers in the most consequential technological shift of this generation. Compliance restrictions, fragmented data lakes, and legacy infrastructure create a structural roadblock to AI adoption even when the intent exists. India’s Digital Public Infrastructure (DPI) is changing the equation. What began with Aadhaar, a national biometric identity layer, has evolved into a full-fledged digital ecosystem: UPI processing over 20 billion transactions in a month, DigiLocker storing verified documents for ~700 million users, Account Aggregator framework enabling consent-based data sharing across 1000+ institutions. What we are witnessing right now may be India’s most underappreciated strategic advantage, it built the infrastructure rails before the shift to AI began, and that sequencing might end up being the biggest changemaker.

“India’s most underappreciated strategic advantage: it built the infrastructure rails before the AI race began.”

The AI Promise

AI investments have been surging constantly over the past decade, yet the results remain elusive. Despite widespread usage, organizations are not seeing the results on their balance sheets, largely because of the cosmetic use cases rather than the core ones. The top use cases currently are post-sales support, chatbots, and marketing. A top reason cited by CXOs for AI currently not used for core use cases is the possibility of false positives, which even if present marginally can significantly erode customer trust. Other challenges that organizations face are regulatory & privacy concerns, legacy infrastructure, and fragmented governance structures which is leading to low output compared to actual investments. Nearly 96% of the organizations have AI adoption planned or deployed but just 4% of them have realized value from their investments (TCS Global AI Study). In BFSI, AI is not constrained by model sophistication or data compute power, but by the availability of trusted, interoperable real-time financial data.

BY THE NUMBERS

Nearly 96% of organizations have AI adoption planned or deployed, but just 4% of them have realized value from their investments. In BFSI, AI is not constrained by model sophistication or data compute power, but by the availability of trusted, interoperable real-time financial data.

India’s Digital Infrastructure

Most markets started chasing AI before building the foundation. India did the opposite, it spent its last decade building a population scale digital infrastructure: identity rails with Aadhaar, payment rails with UPI, data-sharing rails with AA framework, and now faster credit assessment and disbursal using ULI. India’s DPI architecture addresses several of the structural bottlenecks at an ecosystem scale- standardizing governance, simplifying compliance, and enabling interoperability across institutions that previously operated in silos. While others raced to ride the AI wave, India had spent a decade laying tracks.

Each layer of India’s Digital Public Infrastructure removes a structural barrier to AI adoption in BFSI. (AI generated visual)

The Flywheel Begins

DPIs are accelerating financial formalization, pulling people into documented, data-generating financial relationships. AI becomes essential to process this scale automating and augmenting various tasks such as credit underwriting, fraud detection, AML monitoring, and making compliance continuous rather than periodic. The DPI layer supplies data that trains the models, better models drive deeper adoption creating a self-reinforcing flywheel that compounds over time. The only condition here is: companies need to move AI from peripheral to core functions. The direction is right but the pace of adoption in core functions needs to accelerate.

Ecosystem creating a self-reinforcing loop

DPI as Economic Infrastructure for AI

AI effectiveness is fundamentally a data quality problem and in BFSI, data quality has historically been the most persistent issue. Siloed data, inconsistent formats, and absent consent frameworks for accessing customer data have made centralized, technology-led decision-making difficult. DPIs solve this by standardizing data formats across institutions and facilitating consent-led data exchange ensuring interoperability and legal defensibility of data flow. It eliminates one of Enterprise AIs most underestimated costs: the time and expense of acquiring, cleaning, and validating trustworthy data.

DPI has also fundamentally altered how trust is established with customers. Verification processes that once took days of manual effort, identity checks, document validation, etc. can now be completed in seconds through Aadhaar based eKYC and eSign. AI systems can then automate the downstream workflows because the trust infrastructure already exists.

“A customer’s CIBIL score might tell you who they were, but not who they are. UPI’s transaction velocity changes the game.”

In most economies, AI credit models rely solely on static financial records, a customer’s CIBIL score, generally updated quarterly, might tell you who they were but not who they are. UPIs transaction velocity changes the game. With hundreds of micro-transactions generating continuous behavioral signals, institutions can build dynamic credit models, detect fraud in real time, and personalize the offerings to a very granular level.

From Data Owners to Intelligent Operators

The competitive core of the sector is shifting. Owning proprietary data was once the moat. As shared data rails under AA & ULI frameworks mature, the advantage moves to whoever builds the best intelligence over the shared infrastructure. The next frontier for BFSI may be AI-Native financial institutions, banks with autonomous underwriting engines, conversational finance tools augmenting traditional relationship management, and predictive compliance in place of reactive ones. Realizing these requires coordinated action from all stakeholders. Banks and NBFCs must integrate with ULI, AA, and OCEN networks, not as compliance exercises, but as strategic infrastructure decisions. Simultaneously, the institutions will be required to modernize their existing legacy infrastructure to consume and act on real-time signals.

WHAT THIS MEANS FOR BFSI LEADERS

The competitive advantage in banking is shifting from data ownership to intelligence built on shared data rails. Institutions that treat ULI, AA, and OCEN onboarding as strategic decisions, not compliance exercises, will define the next decade of financial services in India.

The Governance Gap

While DPIs & AI can aid the process by making it faster and more transparent for banks and customers alike. It will also amplify systemic risks if the governance mechanisms do not evolve at the same pace as the AI capabilities and the DPI’s adoption does. Specific failure types demand attention: AI hallucinations in credit decisions, embedded bias in fraud monitoring and underwriting tools, and the absence of explainability in outcomes that can end up affecting livelihoods. In a regulated industry, a black-box AI decision is not just a technical issue, it’s also a legal and ethical one. At the same time, the consent frameworks need to be robust and must not create fatigue among the customers. If the users are bombarded with data-sharing requests they don’t understand, they will disengage leading to the same issues of opacity that it was set to solve in the first place.

India’s Moment: If It Moves Fast Enough

India’s DPI stack is not merely a policy achievement; it is a strategic advantage in the global AI race. While other markets scramble to build consent and data sharing frameworks, India already has them. The transaction volumes are real. The regulatory intent exists. What remains now is execution; BFSI institutions moving AI from periphery to the core, use cases evolving to justify the rising investments, regulators building frameworks to keep pace with the rising capabilities, and the whole ecosystem evolving to see DPI as a strategic moat and not just a compliance layer.

“The question is not whether the transformation will happen. It is whether India’s institutions will lead it.”

Just as UPI became the global benchmark for real-time payments, India has everything in place to pull off the AI adoption at the national level. The question is not whether the transformation will happen. It is whether India’s institutions will lead it or sit on the siderails while the whole world leads ahead.