.png){kind=logo}
HDFC Bank, Kotak Mahindra Bank and SBI Life show how Agentic AI has moved from experiments into real BFSI workflows.
Credit, underwriting, fraud investigation, claims and customer service offer some of the strongest opportunities for AI agents.
RBI’s evolving framework places strong emphasis on human oversight, accountability, explainability, security and model risk controls.
India’s banking, financial services and insurance sector has moved beyond basic chatbots and simple automation. Agentic AI now has a place in credit, underwriting, fraud checks, customer service, claims and internal operations. Unlike a conventional chatbot, an AI agent can understand a task, collect information, use digital tools, follow several steps and send complex cases to a human employee.
The shift has strong business value. EY data show that 74% of Indian financial firms had started GenAI proof-of-concept projects, while 11% had already moved GenAI into production. Around 42% had also allocated budgets for GenAI. EY estimates that financial services could gain 34–38% productivity by 2030, while banking operations could see productivity gains of up to 46%.
Indian banks now focus more on workflow automation than simple customer conversations. An AI agent can collect customer documents, read financial information, check internal rules, review credit data, prepare a recommendation and send an unusual case to a credit officer. Such a process can cut manual work while keeping a human employee in control of important decisions.
HDFC Bank provides one of the clearest examples. Its April 2026 disclosure reported five AI-agent use cases in production and another 14 under development. The bank has built an internal AI platform with Agentic Studio and Agentic Mesh, along with enterprise search, document extraction, voice agents, model evaluation and governance tools.
HDFC Bank also reported 97% digital adoption for payments and service transactions and 92% digital adoption for acquisition journeys. This digital base gives the bank more scope to use AI inside complex processes.
Kotak Mahindra Bank has taken a more focused route. The bank has built around 20–25 standalone AI agents for areas such as credit-bureau analysis, document intelligence, OCR, financial-statement analysis and employee support.
One agent can review a credit-bureau report of about 20 pages and create a useful summary for a credit analyst. Kotak keeps human employees in the decision loop, which suits the high-risk nature of financial services.
Credit underwriting stands among the most valuable areas for Agentic AI. A credit agent can collect documents, examine bank statements, check bureau information, compare details with lending rules and prepare an initial credit view. It can also spot missing information and send unusual cases to a human officer.
This model can help banks reduce the time needed for loan assessment. It can also help lenders study customers with limited formal credit records. RBI Governor Sanjay Malhotra recently said AI could help lenders use information such as cash flows, GST filings and utility-payment records. Such data could expand access to formal credit for customers who do not fit traditional credit models.
Also Read - Generative AI and Regulatory Compliance in the BFSI Sector
Insurance offers another major opportunity for Agentic AI. Underwriting and claims contain large amounts of documents, rules and risk information. An AI agent can read medical reports, extract important facts, check policy rules, identify missing details and prepare an underwriting recommendation.
SBI Life has adopted Datamatics’ TruAI Underwriting, an Agentic AI solution for risk assessment. The system can study medical histories, laboratory reports and several risk indicators. The use case shows how Indian insurers have started to take AI into core business decisions rather than limiting it to customer support.
ICICI Lombard also has a broad AI strategy across underwriting, claims and customer service. Its multilingual Voice AI platform has handled more than one million customer conversations. Axis Max Life has explored Agentic AI for underwriting, servicing and IT operations. EY estimates that insurance customer service could gain around 48% efficiency from AI.
Fraud detection and anti-money-laundering work can also benefit from AI agents. A traditional system may flag a suspicious transaction, while an agent can take the next steps. It can collect customer history, review related transactions, compare several alerts, gather evidence and prepare a case for an investigator.
An India-focused research project called CASE tested an AI agent with Google Pay India data and reported a 21% increase in scam-enforcement volume. The result shows the wider value of agentic systems: an agent can help gather information and support investigation rather than simply label a transaction as risky.
The rapid rise of Agentic AI also creates new risks. RBI’s FREE-AI framework, released in August 2025, sets seven principles and 26 recommendations across infrastructure, policy, capacity, governance, protection and assurance. The framework stresses fairness, accountability, explainability, safety and human control.
RBI’s June 2026 draft Model Risk Management guidance adds stronger requirements for AI models. Banks need clear records that support traceability, reproducibility and auditability. The guidance also highlights cyber risks, prompt injection, unauthorized access, external interfaces and unusual system use. Customer-facing AI should identify itself as AI, explain its limits and provide access to human assistance.
Why this Matters
Agentic AI can change how India’s BFSI sector handles credit, fraud, underwriting, claims and customer service. It can reduce manual effort, speed up decisions and widen financial access. Its impact could extend beyond cost savings, with AI helping banks and insurers serve rural customers, MSMEs and people with limited credit histories.
India’s BFSI sector now stands at an important point. GenAI has moved into real business use, while specialised AI agents have started to handle parts of credit, underwriting, claims, fraud and service workflows. Yet fully autonomous financial decisions remain uncommon.
The strongest model for India looks more like bounded autonomy than complete independence. AI agents can handle routine work, analyse large amounts of information and prepare decisions, while human employees retain control over high-risk outcomes.
The biggest opportunity may come from rural customers, farmers, MSMEs and people with thin credit histories. SBI Chairman C.S. Setty recently highlighted AI’s potential for farm credit, alternative data, satellite information and rural financial access. RBI has also compared AI’s potential impact on lending with UPI’s impact on payments.
India’s next BFSI AI phase will therefore focus less on impressive chatbots and more on AI agents that can complete real financial tasks. Banks and insurers that combine trusted data, strong digital infrastructure, secure system access and clear human oversight could gain the largest advantage from this shift.
1. What is Agentic AI in BFSI?
Agentic AI refers to AI systems that can understand tasks, access relevant information, use digital tools and complete several steps with limited human intervention.
2. How are Indian banks using Agentic AI?
Banks use AI agents for credit analysis, document review, customer service, fraud investigation, employee assistance and other operational workflows.
3. Which Indian banks have adopted Agentic AI?
HDFC Bank and Kotak Mahindra Bank have publicly disclosed multiple AI-agent use cases across banking operations.
4. How is Agentic AI changing insurance?
Insurers can use AI agents for underwriting, claims, document analysis, customer service and risk assessment. SBI Life has already adopted an Agentic AI underwriting solution.
5. What are the main risks of Agentic AI in BFSI?
Key risks include inaccurate decisions, cybersecurity threats, data misuse, model risk, prompt injection and weak accountability, which makes human oversight essential.
