SBI Chairman CS Setty Outlines Bank Wide AI Push Spanning Fraud Detection, Lending And Rural Credit Access
India’s largest lender is pushing artificial intelligence deep into fraud detection, credit underwriting and cheque processing, with its chairman now arguing the technology’s real test lies in reaching rural India and small businesses, not just retail convenience.
Highlights:
- SBI used AI to underwrite nearly Rs 1 trillion in MSME loans during FY26
- Chairman CS Setty says the bank’s entire cheque processing is moving to AI
- SBI’s cost to income ratio has improved to around 51.6 percent
- The bank has automated more than 220 back end processes using AI
- AI generated business already contributed roughly Rs 22,000 crore in Q1 FY27
- Setty wants AI’s next phase focused on rural India and small business credit
When the chairman of a country’s largest bank stands up at a major industry conference and says the real test of a technology will not be how sophisticated it becomes, but what it actually enables people to accomplish, it is worth paying attention, both to the sentiment and to what sits behind it. That is precisely how CS Setty, Chairman of the State Bank of India, framed the bank’s artificial intelligence strategy at the FIBAC 2026 annual banking conference in Mumbai this month, and the numbers backing up that statement suggest this is considerably more than aspirational language layered onto business as usual.
SBI’s AI push is not a single initiative confined to one department, it spans what Setty himself described as the entire customer lifecycle—credit underwriting, portfolio management, fraud risk management, and customer service—deployed across an institution that remains, by a wide margin, India’s largest bank by assets and customer base. Some of the concrete results already on record are genuinely substantial. SBI Managing Director Rama Mohan Rao Amara disclosed that the bank used AI to underwrite nearly Rs 1 trillion in MSME loans of up to Rs 5 crore each during the financial year 2025 to 2026, spanning both new-to-bank customers and existing ones. That underwriting process combines data from the goods and services tax network, GST filings, account information, credit bureau scores, and other unstructured data sources, processed through what the bank calls its Business Rule Engine, a system designed specifically to speed up lending decisions that would otherwise require considerably more manual analysis from relationship managers.
The operational side of this transformation is arguably where the most tangible near-term impact has shown up. Setty confirmed that SBI’s entire cheque processing operation is being transitioned to AI, building on an existing system that already handles cheques valued at Rs 10,000 or less, which account for around a quarter of the bank’s total cheque volume, through fully automated straight-through processing with practically no human intervention.
“Our total cheque processing is moving to AI, it releases a lot of workforce which is currently employed in manually doing it.”
The bank has retained a control mechanism throughout this shift, with a dedicated risk unit reviewing a sample of AI-processed cheques specifically to catch errors and determine when underlying models need further training, a reasonable safeguard given the scale of transaction volume now flowing through largely automated systems.
Beyond lending and back-office processing, SBI has been building AI into its fraud detection and cybersecurity infrastructure as a core priority rather than a peripheral experiment. The bank uses AI-based early warning signals to identify vulnerable loan exposures before delinquencies actually emerge, drawing on sector-specific market data and other publicly available information to flag risk earlier than traditional monitoring would typically catch it. At its security operations centre, AI systems process the enormous volume of logs generated across the bank’s IT infrastructure to detect potential threats, while a separate resiliency operations centre uses AI to identify and predict possible system breakdowns before they actually occur. According to figures cited in independent academic analysis of SBI’s AI strategy, the bank has automated more than 220 distinct back-end processes altogether, a scale of automation that has coincided with a cost-to-income ratio improving to around 51.6 percent and a gross non-performing asset ratio holding at a comparatively low 1.82 percent, both figures that speak to genuine operational discipline rather than automation pursued purely for its own sake.
The commercial payoff from all of this has also started showing up directly in SBI’s quarterly financial disclosures, not merely in efficiency metrics. The bank posted a record Q1 FY27 consolidated net profit of Rs 24,113 crore, and disclosed separately that Rs 22,000 crore worth of business had already come through AI and analytics generated leads across home loans, gold loans, and MSME lending specifically. That figure deserves an honest caveat rather than uncritical repetition: the Rs 22,000 crore represents business generated through AI-assisted leads rather than incremental profit created purely by AI, and it remains genuinely difficult to determine how much of that business might have materialised through traditional channels regardless, a distinction that matters for anyone trying to assess AI’s true incremental contribution to the bank’s bottom line rather than simply its role in an already growing loan book.
What distinguishes Setty’s recent public remarks from a fairly standard corporate AI adoption narrative is the specific direction he has pointed toward for the technology’s next phase. Rather than treating AI’s current retail banking successes as the primary achievement to build on, Setty argued explicitly that the next wave of AI-led banking in India must move deeper into the economy, toward rural India, small businesses, and customers whose financial histories do not fit conventional credit models. He specifically flagged agriculture as an area of significant opportunity, pointing to how data-driven risk assessment, digital farm records, and satellite imagery could collectively help banks improve both credit access and portfolio management in a sector that has historically been underserved by formal lending infrastructure.
“The challenge is to take these capabilities beyond pilots and make them affordable, practical and accessible at the last mile.”
Setty was also careful not to present this expansion as risk-free. He specifically flagged that greater reliance on AI would introduce new categories of vulnerability, including more sophisticated cyber threats and fraud techniques capable of evolving faster than conventional defensive systems can adapt, meaning banks expanding AI usage would simultaneously need to strengthen their underlying security posture rather than treating AI adoption and cybersecurity investment as separate workstreams. He also emphasised the importance of human accountability scaling in proportion to how consequential a given AI-driven decision becomes, a governance principle that will likely face genuine tests as SBI pushes AI further into higher-stakes lending decisions affecting rural and small business borrowers who may have far less recourse or financial literacy to challenge an automated decision than a typical urban retail banking customer would.
Viewed as a whole, SBI’s AI transformation reads as considerably more substantive than typical corporate technology messaging, backed by concrete underwriting volumes, verified efficiency metrics, and a leadership team willing to acknowledge both what has worked and what remains genuinely unresolved. At the same time, the harder test Setty himself has explicitly set—extending AI’s benefits meaningfully into rural India and small business lending rather than concentrating gains primarily among SBI’s already digitally engaged retail customers—remains a considerably more difficult undertaking than the retail and back-office automation successes achieved so far, one that will require sustained investment in last-mile infrastructure, financial literacy, and careful governance well beyond what has been required to automate cheque processing or accelerate urban MSME underwriting.
