‘Know Your Agent’ needed as AI enters banking; 99% accuracy is not enough: SBI Chairman C.S. Setty

State Bank of India (SBI) Chairman C.S. Setty on Thursday called for a new framework to identify and authenticate artificial intelligence agents as banks move towards autonomous systems, saying the financial sector may increasingly need to shift from “Know Your Customer” to “Know Your Agent”.

Speaking at the Global Fintech Fest 2026 in Mumbai, Setty said the transition from AI being used as a tool to technology capable of acting independently on behalf of customers would fundamentally change the risk landscape for banks. “The moment AI moves from recommendation to execution, accountability becomes even more important,” he said.

AI moves from assistance to execution

Setty said agentic AI could operate across the financial lifecycle, including fraud and new-account detection, KYC and anti-money laundering processes, loan appraisal and underwriting, reconciliation and customer lifecycle management.

SBI, he said, has already been an early adopter of AI and machine learning, with applications spanning customer-service chatbots, pre-approved personal lending, credit assessment, cash-flow-based lending, early-warning systems, fraud detection, AML monitoring and proactive risk management.

However, the next generation of AI would be different, Setty said, as agents could move beyond performing defined tasks to responding to changing circumstances and acting on behalf of customers. At India’s scale, he said, this could improve decision quality, reduce response times, strengthen controls and make the financial system more resilient.

‘99% accuracy is not enough’

Setty said trust must be embedded into agentic AI rather than added after deployment, warning that an autonomous system could trigger a chain of actions across multiple systems if it makes an error.

“There is nothing like 99% accuracy. You need to be 100% accurate, and every time, all the time,” Setty said, adding that average accuracy was not sufficient when AI operates at population scale.

He identified three pillars of “trust at scale” — accuracy, accountability and access without asymmetry. AI systems, he said, must perform consistently across customer segments, while important decisions should have monitoring, traceability and mechanisms to establish why an action was taken.

Setty also stressed the need for a “human in the loop”. While routine, high-volume and data-intensive activities could be handled by AI, complex and high-risk financial decisions should continue to receive appropriate human oversight.

From KYC to ‘Know Your Agent’

Setty said banks had spent decades building robust processes around knowing their customers, but autonomous AI agents participating in financial transactions would create a new requirement: knowing the agent acting on a customer’s behalf.

Such a framework, he said, would need to cover agent identity, authentication, consent, transaction limits, audit trails and location. He also warned of an emerging “AI versus AI” environment, requiring banks to combine identity intelligence, behavioural intelligence, transaction intelligence and real-time risk assessment.

Setty said India was uniquely positioned for the next phase, having already demonstrated that digital financial infrastructure could serve a billion people. The opportunity now, he said, was to move from digital inclusion to “intelligent inclusion”, with affordable, scalable and India-specific AI making financial services more accessible, relevant and responsive to hundreds of millions of customers.