SBI wants to put AI across the bank with $1 billion investment
SBI is moving beyond isolated AI pilots to build an integrated AI strategy across banking, with governance and process changes at its core.
India's largest bank is preparing to take AI from individual experiments to a bank-wide system. State Bank of India (SBI) plans to invest about $1 billion in building an enterprise-wide AI framework that will connect artificial intelligence across customer services, operations, risk and other functions.
SBI CIO Abhay Pandey outlined the plan at the ETBFSI CXO Conclave in Mumbai. This shows a shift from running separate AI pilots to creating a common architecture that can support AI across the organisation.
Why SBI wants one AI framework
SBI's scale gives it access to huge volumes of customer and operational data.
According to Pandey, that scale makes a fragmented approach to AI less effective and creates a need for a common platform. The bank has already begun the process and issued requests for proposals, suggesting the programme is moving towards implementation.
The proposed framework is expected to connect customer-facing applications with back-office functions. This could allow AI systems to work with a wider range of information instead of keeping individual use cases isolated within different departments.
For a bank of SBI's size, the potential applications extend well beyond customer interactions. AI could support fraud detection, compliance, risk management, operational efficiency and cost optimisation.
Governance will be central to the rollout
SBI's AI strategy also places strong emphasis on governance and risk controls. Pandey highlighted the need for a new governance structure to manage risks associated with AI models. One concern is hallucination, where an AI system generates information that sounds convincing but is incorrect.
For a financial institution, such errors can have serious consequences. AI outputs therefore need appropriate checks, oversight and human intervention before they are used for important decisions.
SBI also plans to bring structured and unstructured data into the wider AI framework. Structured data is information organised in formats such as databases, while unstructured data includes documents, emails and other content that does not follow a fixed format.
AI will need process changes too
The bank is pairing its AI investment with a broader process reengineering exercise. This is important because simply adding AI to existing workflows does not automatically make them more efficient. SBI will need to redesign processes so that employees and AI systems can work together effectively.
The approach could help the bank reduce repetitive work while allowing employees to focus on tasks that require judgement and oversight. The next stage will be important. Contract awards from the current RFPs, details of the governance framework and measurable results from early deployments will show how quickly SBI can turn the $1 billion plan into operational improvements.
If the strategy works, SBI could provide a blueprint for how large financial institutions can deploy AI at scale while balancing efficiency with control.