Bank of India is stepping up its artificial intelligence and analytics push with investments across data infrastructure, AI-ML models, automation, and cybersecurity, as the lender looks to improve operational efficiency, underwriting quality and risk management.
According to Rajneesh Karnatak, speaking during the post FY26 results announcement, “the bank allocated nearly Rs 2,000 crore towards IT-related expenditure in FY26”, covering digital initiatives and cybersecurity. More than 80% of the allocation was spent through a combination of capex and opex during the year.”
On cybersecurity, Karnatak said the bank had already spent a significant portion of its FY26 IT allocation towards strengthening cyber resilience. “Nearly 9% of the total IT allocation was earmarked for cybersecurity, of which about 80% of the budgeted amount was utilised.”
Given the increasing cybersecurity threats facing the banking sector, “the bank plans to increase its cybersecurity budget by nearly 35% in FY27,” he added.
The bank has also strengthened operational monitoring through its Resiliency Operations Center, which runs a 24x7 monitoring setup for critical banking applications. “This helped improve the Mean Time to Detect (MTTD) from 45 minutes to 15 minutes, enabling faster incident resolution,” according to the Bank.
Karnatak said automation initiatives under ‘Project Star Aditya’ helped the bank, “save over 68,000 working hours through 48 automation initiatives undertaken across the organisation. He added that 19% of the loan sanctions during FY26 were completed entirely through digital platforms.”
According to Bank of India, Project Star Aditya currently has 35 live use cases that generated business worth Rs 11,419 crore during FY26.
Bank of India’s AI Playbook - ‘Project Star Aditya’
The broader AI and analytics strategy, however, is being driven through a structured data and AI roadmap that started nearly one-and-a-half years ago, according to Satyendra Singh, Chief General Manager and CIO, Bank of India, speaking exclusively with Financial Express FUTECH.
Singh said the bank realised the importance of 'data, automation, analytics, AI and ML’ and decided to first consolidate and clean the data by building a data lake architecture.
“One and half years ago, we had a data warehouse. The bank realised the potential of data, automation, analytics, AI and ML. The idea was first to clean and bring the data to one place. Following this, we implemented the data lake,” Singh said.
The bank subsequently built an AI-ML practice layer over the data lake, using a combination of open-source technologies to ensure access to established AI-ML libraries, ecosystems and applications.
Singh said the bank has developed the capability to, “ingest semi-structured, structured and unstructured data using object storage infrastructure. While the bank continues to use a traditional data warehouse for reporting requirements, its AI-ML initiatives are running on a broader analytics architecture.
“For reporting, we are using Qlik. It has an inbuilt AI capability, but still its role is limited to interactive dashboard,” he said.
The AI-ML layer is largely dependent on open-source technologies, with R and Python forming the foundational stack for model development.
The bank also simultaneously invested in talent acquisition, onboarding data scientists, data engineers and statisticians, while working alongside Accenture to operationalise the roadmap. “At present, we have 35 AI-ML models, which are deployed,” Singh said. The models are being integrated into the bank’s digital underwriting systems, where automated verification, scoring and appraisal mechanisms assist credit sanctioning processes.
According to Singh, the AI-led underwriting and monitoring systems have contributed to reducing delinquency levels and improving risk assessment capabilities.
“One of the very good use cases is where we are predicting that even if an account is 100% healthy today, what are the chances in this financial year that it will go delinquent,” he said. The system analyses behavioural and segment-based patterns and sends alerts to field teams wherever risks begin emerging, enabling focused intervention before slippages occur.
GenAI Usecases
With its AI-ML foundation now stabilising, the bank is moving towards generative AI and agentic AI capabilities as part of the next phase of its roadmap.
According to Singh, the bank has adopted a hybrid infrastructure strategy for generative AI deployments because of regulatory and operational constraints around full-scale cloud adoption. “We cannot do 100% on cloud. So for the generative AI and agentic AI part, we have taken a call for a hybrid kind of setup,” he said.
The bank has started experimenting with large language models (LLMs), including India-based models, and is currently refining generative AI use cases around proposal summarisation for mid-sized credit proposals. The objective is to reduce the time taken in preparing, analysing and summarising proposals while identifying key points requiring resolution.
“We wanted to do it with a good business use case,” Singh said, adding that the bank is attempting to build its own intellectual property around these capabilities.
While the AI-ML investments were largely undertaken during FY25-26, the bank plans to invest a similar amount in FY27 towards generative AI initiatives. Singh estimated that, “the AI-ML layer itself involved investments of roughly Rs 30-60 crore (FY26), while a similar amount is expected to be allocated for generative AI capabilities going forward (FY27). GPUs are costly. We are still evaluating the optimal mix between on-premise infrastructure and cloud components within a hybrid deployment model,” he said.
