• AI deployment: Bankers described AI as a journey from pilots to implementation, with Indian banks well positioned to develop use cases. Rajiv Anand said banks should redesign customer journeys—not simply add AI to existing processes—and cited around 300, 400 internal agents.
• Credit productivity: AI can analyze unstructured information, such as shop images and customer calls, and speed up loan assessment using bank statements, GST data and other records. One bank handles close to 1 million loans a month, or 10 lakh, and runs 150 machine-learning models; Subramaniakumar said AI could raise a credit team’s capacity from around 100 proposals a day by 5x or 10x, depending on deployment.
• Controls and accountability: Speakers emphasized security, guardrails, bias and hallucination prevention, explainability, traceability and resilience. Banks remain accountable for AI-supported decisions, and agents must be prevented from accessing unauthorized data.
• Investment and returns: Vaidyanathan said banks invest about 10% of operating expenditure in technology, while Anand described technology spending as generally optimal. Anand cautioned that efficiency gains may ultimately be competed away to customers through faster, cheaper services; Vaidyanathan said AI was contributing to lower costs, with his bank’s incremental cost-to-income ratio at 55%.