Winning in the AI Era – Indian Banks

 

FICCI – FIBAC Compendium - 20263

Winning in the AI Era – Indian Banks

R Kannan

Introduction

The Indian banking sector stands at a historic inflection point, moving beyond traditional digitization into the era of Artificial Intelligence (AI). Supported by a robust macroeconomic foundation, historic lows in gross non-performing assets, and a world-leading Digital Public Infrastructure (DPI), Indian banks are uniquely positioned to redefine customer engagement, risk management, and financial inclusion. However, as institutions transition from small-scale AI pilots to enterprise-wide adoption, success will not depend solely on technological sophistication. Long-term leadership will belong to banks that effectively combine machine intelligence with human judgment, robust governance, operational resilience, and unwavering customer trust.

Observations from the Report

Credit and Deposit Growth Acceleration

Indian bank credit growth reached 14.5% year-on-year in FY26, exceeding the 11.0% growth recorded in FY25.

Deposit growth similarly expanded, rising from 10.3% in FY25 to 11.5% in FY26.

The national credit-to-GDP ratio improved noticeably from 51.6% in 2020 to 61.8% by 2026.

These metrics demonstrate the sector's structural resilience amid global trade tarif3fs and geopolitical volatility.

Substantial Gains in Capital Adequacy and Asset Quality

Scheduled commercial banks' Capital Adequacy Ratio (CAR) improved to 17.68% in March 2026 from 17.10% in 2023.

Gross NPAs decreased dramatically from 3.9% in March 2023 down to a multi-decadal low of 1.73% in March 2026.

Net NPAs dropped to 0.40% in March 2026, compared to 1.0% recorded in March 2023.

Combined net profit of listed public and private sector banks reached ₹3.94 lakh crore in March 2026.

Addressing the Operational Expense and Productivity Gap

Despite extensive digital transformation over 15 years, real bank productivity gains have averaged only ~1% annually.

Operational expenditure ratios continue to rise as operating expenses outpace operating income growth.

Mature AI deployment offers the capability to automate 35–40% of low-value, routine back-office tasks.

Closing this gap requires shifting from isolated AI pilots to enterprise-wide production deployment strategies.

Strategic Adoption of Modular "Surround-and-Extend" Architectures

Banks are actively transitioning away from rigid, monolithic legacy technology stacks to API-first architectures.

Institutions favour a "surround-and-extend" model that wraps middleware around core legacy systems.

This approach allows for step-by-step AI deployment without causing costly disruptions to daily banking operations.

It enables real-time integration of advanced AI capabilities alongside existing legacy software setups.

Adherence to the RBI FREE-AI Governance Framework

Banks align ethical AI deployment with the Reserve Bank of India’s FREE-AI Framework guidelines.

Core controls mandate a "Human-in-the-Loop" for critical decisions like credit underwriting.

Technical "Model Kill Switches" are required to instantly halt systems exhibiting algorithmic bias or error.

Data collection practices must strictly align with standards set by the Digital Personal Data Protection Act.

Mitigating AI-Accelerated Security Risks and Vendor Dependency

Generative AI is being weaponized by bad actors to automate zero-day exploits and run advanced phishing campaigns.

Deepfake-related fraud in India has increased by over 550% since 2019, targeting financial systems.

The complex "Black Box" nature of machine learning hinders the explainability of high-stakes credit choices.

Regulators place absolute legal responsibility on banks regardless of third-party tech vendor reliance.

Projections for AI Economic Impact and National Trust Advantage

AI is estimated to contribute over USD 500 billion to the broader Indian economy by the year 2030.

Studies indicate 73% of Indian business leaders expect India to emerge as a leading global AI nation by 2030.

Traditional banks hold a global trust score of 68%, surpassing fintechs (53%) and crypto firms (43%).

Preserving this core trust advantage remains vital as decision-making algorithms become more automated.

Scale and Success of Digital Public Infrastructure (DPI)

India’s DPI stack, including UPI, Aadhaar, DigiLocker, and Account Aggregator, is a global benchmark.

In FY2025-26, UPI processed over 24,162 crore transactions with a total value exceeding ₹314 lakh crore.

More than 700 banks are currently live and processing real-time payments on the UPI network.

DPI creates a consent-based, interoperable foundation that feeds verified data directly into AI algorithms.

Data as the Primary Constraint to Scaling AI Enterprise-Wide

Studies show that only 7% of organizations globally have fully scaled AI implementations across the enterprise.

Data fragmentation across siloed banking systems serves as the main bottleneck to scaling intelligent systems.

Competitive advantage stems from organizing data into a traceable, governed, and unified enterprise foundation.

Modernizing data infrastructure yields far greater long-term ROI than merely running high-profile technology pilots.

Focusing on Process Augmentation over Mere Automation

While automation cuts costs, the primary economic upside comes from augmenting human judgment with AI.

Augmentation improves risk accuracy, accelerates fraud detection, and surfaces early market opportunities.

Combining human prudence and contextual understanding with machine speed yields optimal financial outcomes.

Banks using agentic operating models and AI at scale can potentially achieve net cost reductions of 15–20%.

Narrowing the MSME Credit Gap via Alternative Data

AI-driven models can unlock USD 130–170 billion in economic value by bridging the underserved MSME credit gap.

Models leverage Account Aggregator flows, GST records, and cash-flow patterns instead of physical collateral.

Evaluating real-time business health enables banks to extend credit safely to thin-file and first-time borrowers.

Shift toward cash-flow-based underwriting enhances financial inclusion across previously excluded segments.

Transitioning to Real-Time Predictive Risk and Asset Tokenization

Advanced analytics shift risk management from reactive post-mortems to real-time predictive monitoring.

AI detects stress signals—such as irregular repayments and cash-flow drops—long before default occurs.

Tokenization of collateral and receivables onto programmable rails provides verifiable visibility into asset health.

AI models will predict credit risk directly from asset movement rails, enabling proactive NPA prevention.

Language Inclusivity via Conversational Voice Banking

India has over 958 million active internet users, with 57% residing in rural regions where local languages dominate.

Voice-enabled, regional-language AI assistants allow customers to conduct banking using native speech.

Conversational interfaces lower literacy, geographical, and technological barriers for rural customer segments.

Multilingual language models provide round-the-clock service while eliminating regional communication gaps.

Strategic Phased Adoption Framework for Intelligent Banks

Transformation must follow three phases: Build (data/governance foundation), Integrate (workflows), and Reinvent (models).

AI projects fail to create enterprise value when confined solely to isolated IT proof-of-concepts.

Operating models must re-align around embedded intelligence in every transaction, decision, and workflow.

Long-term reinvention involves pairing AI with frontier tech like blockchain to solve industry-wide challenges.

Enterprise-Wide AI Deployment Examples at Bank of Baroda

Bank of Baroda leverages an 8-petabyte Enterprise Data Lake to run over 60 active AI/ML production use cases.

Deployed platforms include ADI (virtual assistant) and ADITI (multilingual Virtual Relationship Manager).

Employee productivity and customer support are augmented via GyanSahay (GenAI knowledge platform) and SAMVAAD.

Applications span digital lending, wealth management, collections, treasury, and automated fraud monitoring.

Implementation of Industry-Wide Anti-Fraud Tools Like MuleHunter.AI

The RBI Innovation Hub deployed MuleHunter.AI to proactively detect fraudulent mule account networks across banks.

Graph Neural Networks map complex transaction relationships to uncover hidden, coordinated fraud rings.

Automated rule engines monitor transactions continuously to contain cross-channel fraud in real time.

Collaborative, industry-wide intelligence platforms help neutralize machine-speed financial crimes.

Model Risk and the "Right to Explanation" Requirement

Black-box algorithms pose significant regulatory liabilities when customers are denied credit without clear reasons.

Borrowers possess a regulatory right to know the precise basis behind an adverse credit decision.

Explainable AI (XAI) models ensure credit algorithms operate transparently, fairly, and free from historical bias.

Models require continuous monitoring to catch data drift and prevent discriminatory scoring against vulnerable groups.

The Strategic Advantage of Public Sector Banks in Inclusion

Public sector banks possess extensive regional networks well-suited for driving AI-led financial inclusion.

Indigenous language models allow PSBs to economically serve rural traders, farmers, and women-led enterprises.

AI helps bridge long-standing cost barriers associated with serving low-ticket, geographically scattered accounts.

Deep physical presence paired with digital intelligence creates a scalable "phygital" banking model.

Developing Custom AI Models Tailored to Indian Contexts

Imported global AI models frequently fail because they are not tuned to Indian languages, behaviours, or economic conditions.

Leveraging the national IndiaAI Mission allows banks to build localized, highly context-aware models.

Homegrown domain models deliver higher accuracy, lower processing latency, and superior cost economics.

Indian context models enhance underwriting accuracy for unique demographics like self-help groups (SHGs).

Proactive Consumer Protection through Behavioural Nudges

Specialized AI models replace broad awareness campaigns with personalized, contextual consumer alerts.

Algorithms deliver real-time behavioural nudges when a transaction pattern indicates potential fraud exposure.

Personalized product matching ensures product suitability, minimizing the systemic risk of financial mis-selling.

Early intervention models protect vulnerable retail consumers before actual financial harm takes place.

Optimizing Operational Workflows via Generative AI

Large Language Models assist employees in drafting credit notes, summarizing interactions, and analysing policies.

Generative tools extract structured insights from complex documentation, sharply reducing loan turnaround times.

Internal knowledge assistants enable staff to quickly search internal circulars, improving operational consistency.

Administrative tasks are streamlined, freeing banking personnel to focus on high-value client advisory roles.

Constructing Human-Allied Workforce Transformation Strategies

Future success depends on human-machine collaboration rather than replacing employees with software.

Financial institutions must heavily invest in continuous reskilling, AI literacy, and capacity-building programs.

HR functions can utilize ethical AI for internal skill-mapping, talent recruitment, and personalized career training.

Final decisions regarding employee performance, promotions, or hiring must strictly remain with human management.

Data Governance, Hybrid Cloud, and Data Privacy Mandates

Strict controls require keeping personally identifiable information (PII) secure on-premises.

Controlled cloud environments can be safely leveraged for less sensitive workloads and advanced analytics.

Auditability, data loss prevention (DLP), and continuous oversight are required parameters for all AI deployments.

Operational policies must comply with the Digital Personal Data Protection (DPDP) Act of 2023.

Five Core Priorities for Future-Ready Banking Institutions

Establish a consolidated, high-quality, enterprise-wide data foundation as an institutional asset.

Implement governance by design with board-approved policies, model tracking, and independent validations.

Retain humans-in-the-loop for credit decisions, grievance resolutions, and complex underwriting tasks.

Build dedicated institutional capacity through ongoing employee upskilling and technical training.

Develop custom solutions aligned with local infrastructure, regional languages, and Indian market conditions.

Matching AI Model Size to Specific Banking Tasks

Not every banking use case requires massive, compute-heavy Large Language Models (LLMs).

Smaller, domain-specific models frequently deliver superior accuracy, lower latency, and better economics.

Systems must balance processing performance, operational costs, system flexibility, and strict data security.

Choosing right-sized models ensures sustainable financial returns on enterprise technology investments.

Establishing Centralized Closed-Loop Customer Feedback Systems

AI-driven repositories unify unstructured customer feedback from call transcripts, emails, apps, and branches.

Systems must run as actionable closed loops, assigning owners, target dates, and root-cause resolutions.

Systemic insights derived from complaints directly inform product design, staff training, and model refinement.

Strict access controls are mandatory to safeguard sensitive financial data contained within customer feedback.

Democratizing Rural Credit via Alternative and Community Data

Over 190 million Indian adults remain credit-invisible due to a lack of formal bureau history.

AI evaluates psychometrics, satellite imagery of farmland, and group (SHG/JLG) ecosystem dynamics.

Assessing group ecosystem health allows banks to extend micro-loans without relying on traditional credit scoring.

Alternative data brings self-employed individuals and rural micro-enterprises into the formal credit fold.

Strategic Agility as an Advantage for Small Finance Banks (SFBs)

SFBs can leverage shorter decision cycles and close community ties to adopt AI rapidly.

Smaller, well-governed institutions implement process changes without navigating heavy bureaucratic layers.

AI acts as a capability equalizer, giving nimble niche banks tools previously limited to mega-institutions.

Proximity to local markets combined with intelligent automation amplifies the core strengths of SFBs.

Establishing Sector-Wide Shared AI Infrastructure Platforms

Industry bodies and banks can create shared platforms for fraud intelligence, verified datasets, and security alerts.

Federated learning approaches allow shared model training without exposing proprietary raw customer data.

Pooled infrastructure gives smaller institutions access to advanced capabilities without high standalone costs.

Collaborative networks strengthen defences against multi-bank fraud vectors and coordinated cyberattacks.

Board-Level AI Governance and Oversight Imperatives

AI strategy and governance must be driven directly by the Board of Directors, not isolated in the CTO's office.

Boards must actively evaluate forward-looking risk models, lending resilience, and societal impact.

Dedicated AI Ethics Committees should review high-impact models for bias, fairness, and explainability.

Investment funding must be tied to verified business outcomes, with underperforming experiments pruned decisively.

Conclusion

The transformation from digital to intelligent banking marks a fundamental evolution in how financial institutions operate, manage risk, and deliver value. While AI offers tools to streamline workflows, democratize credit, and detect fraud, technology alone cannot replace the foundational pillar of banking: trust. Institutions that succeed in this new era will be those that pair analytical power with robust governance, human empathy, regulatory accountability, and high-quality data foundations. Ultimately, AI is not a replacement for human judgment but its powerful partner, positioning Indian banks to drive sustainable, inclusive economic growth across the nation.