Mumbai: State Bank of India (SBI) used artificial intelligence to underwrite nearly ₹1 trillion of loans to micro, small and medium enterprises (MSMEs) during FY26, Managing Director Rama Mohan Rao Amara said at FIBAC 2026. The loans, each of up to ₹5 crore, covered both new-to-bank and existing customers.
SBI’s AI-led underwriting system combines GST data, credit-bureau scores, bank-account information and other structured and unstructured data to assess borrowers. The technology has reduced the amount of time relationship managers previously spent gathering information and conducting preliminary analysis.
AI helps assess MSME borrowers
According to Amara, the use of AI has also helped improve the quality of the loan portfolio. SBI has seen lower delinquency in the portfolio underwritten through its business rule engine, or BRE.
The bank is also using AI to assess customers with limited credit histories, particularly small businesses and proprietorships. These “thin-file” customers may have limited traditional credit information, but SBI can use account and UPI-related data to assess their creditworthiness.
The approach is aimed at expanding financial inclusion while helping the bank meet its priority-sector lending obligations.
AI identifies vulnerable exposures
SBI is extending AI beyond loan origination to portfolio management and risk monitoring. The bank uses AI-based early-warning systems to identify potentially vulnerable exposures before conventional signs of stress, such as delinquencies or days past due, emerge.
The models can analyse large volumes of information, including market conditions, sector-specific developments and publicly available data, to generate early-warning signals.
This allows the bank to identify potential risks earlier and take preventive action instead of waiting for a borrower to show conventional signs of financial stress.
Cheques up to ₹10,000 automated
SBI is also using AI, including large language models and vision-based AI, to automate cheque processing. Cheques of up to ₹10,000 account for around 25% of the bank’s total cheque volumes.
The AI system can read cheques, verify mandatory fields and check compliance requirements. These transactions are now processed through a straight-through processing system with practically no human intervention.
However, SBI has retained a human control mechanism. A dedicated control risk unit reviews a sample of cheques processed by the AI system to identify errors and determine whether the models require additional training.
Generative AI moves beyond pilot projects
SBI has also begun moving some of its generative AI experiments into full-scale applications. The bank is deploying AI across several areas, including credit underwriting, portfolio management, fraud risk management, customer service and operations.
In customer service, AI has progressed from assisting human agents to handling some customer calls directly, with more complex cases being transferred to employees.
The bank is also exploring an agentic AI assistant for corporate banking that can analyse financial statements and documents and assist with risk assessment.
Focus on efficiency and financial inclusion
Amara said that while SBI may take time to quantify the impact of AI through metrics such as a defined reduction in the cost-to-income ratio, the bank is already seeing benefits in customer satisfaction and employee productivity.
The technology is freeing relationship managers from routine data collection and preliminary analysis, allowing them to focus more on validating AI-generated outputs and maintaining customer relationships.
SBI’s expanding use of AI reflects a broader shift in banking towards technology-led credit assessment, risk monitoring and customer service, while the bank continues to retain human oversight in areas where automated systems could potentially make errors.
