Indian bank chiefs call for AI to move from pilots to governed, large-scale deployment
At the Business Standard BFSI Insight Summit, chief executives from IndusInd Bank, RBL Bank and IDFC First Bank argued that Indian lenders are ahead on AI experimentation but must now commit to controlled, outcome-linked roll-outs. They stressed data governance, board-level AI policy and human accountability for customer-affecting decisions.
Leaders of some of India's largest private banks used the Business Standard BFSI Insight Summit to make a pointed argument: the experimentation phase of AI in banking is over, and the sector now needs to scale what works under proper governance. Speakers included the chief executives of IndusInd Bank, RBL Bank and IDFC First Bank.
IndusInd's R Subramaniakumar said Indian banks had been early and energetic adopters, but that the emphasis must shift from proving the technology to deploying it against defined business outcomes. IDFC First Bank's V Vaidyanathan placed data governance on an equal footing with the technology itself and said every bank should have an explicit AI policy, with human oversight retained wherever a system touches credit, compliance, risk or the customer.
The remarks echo survey data from earlier in the autumn. A Zeta study of 18 banks and NBFCs found that around seven in ten run AI in production, but mostly in low-stakes, reviewable tasks such as customer support, fraud screening, document handling and software testing. Responsible-AI frameworks were under development at most institutions, yet none reported organisation-wide implementation, and only a fifth rated their model-risk management as very mature.
Security was the other recurring concern. Executives noted that as agents gain the ability to act on accounts and systems, the control environment must evolve faster than the use cases, and that regulators will expect banks to demonstrate it.
Why it matters
When the people who sign off on budgets say publicly that AI must be scaled with governance rather than parked in pilots, procurement follows. The next twelve months are likely to see Indian banks consolidating dozens of point experiments into a smaller number of platform decisions.
Indicops view
We have watched this pattern in both banks and government departments: a hundred pilots, none in production, because no one owns the governance layer. Our approach is to build that layer first, with model inventory, evaluation gates, audit logging and escalation paths, so that each subsequent use case is a configuration rather than a fresh risk review. It is less glamorous than another chatbot, and it is what gets AI into the core.
More in Enterprise AI
Back to the newsroomFireworks AI breach exposes Notion AI chat transcripts, its second incident in two months
An attacker who breached Fireworks AI's cloud provider reached a development system and took customer billing data and some datasets, including Notion AI chat transcripts. Notion says about 1 per cent of its customers were affected. Fireworks believes the motive is financial, selling access to models.
PhysicsWallah says AI now writes about 90 per cent of its code as headcount and faculty dip
PhysicsWallah told analysts that AI now writes around 90 per cent of its code, and credited AI efficiencies for slightly lower headcount and faculty numbers in the second quarter of FY27. Profit doubled to Rs 142 crore on revenue of Rs 1,242 crore, while the offline business remains loss-making.
Indian IT stocks hold firm through a global AI rout as oil and Hormuz keep investors on edge
While the Nasdaq suffered its worst day since April 2025 and Asian chip stocks fell 6 to 7 per cent, Indian benchmarks extended a seven-session winning streak on 6 November with Infosys, TCS and HCLTech in focus as a services-led alternative to AI hardware. Brent's swing back towards $100 on Hormuz disruption remains the sector's main macro risk.