GenAI In The Indian BFSI Space: Opportunities, Challenges and Roadmap

While everyone is talking generative artificial intelligence, how are Indian BFSI organisations actually deploying GenAI solutions? And while the opportunities are often spoken about, what are the challenges and especially the guardrails necessary in the critical BFSI space. We speak to Siddhesh Naik, Country Leader, Data, AI & Automation Software, Technology Sales, IBM India & South Asia, to get a perspective.

What are the use cases for which Indian BFSI firms are implementing AI and GenAI?

Siddhesh: Traditional AI has been used to identify patterns, make predictions and carry out specific tasks based on historical data and supervised learning. Now, GenAI is opening up new possibilities for enterprises considering that the speed, scope and scale of this technology is unprecedented—made possible with foundation models trained across the breadth of enterprise data.

A prominent use case for GenAI in BFSI today is customer service or agent assistance, where this technology is harnessed to surface insights and bring relevant information to service agents’ fingertips, so they can respond to customer queries faster and with greater confidence. Increasingly, GenAI is also being leveraged by BFSI organisations in marketing to aid content generation and enable hypepersonalised marketing campaigns. Other popular areas include HR automation, trade finance automation, threat management, regulatory compliance and application modernisation.

Are GenAI implementations mostly at the pilot stage or are Indian BFSI firms seeing commercial deployments of GenAI?

Siddhesh: Many BFSI organisations are still in the pilot stage with GenAI. However, from my discussions with customers, organisations are trying to move internally facing GenAI applications into production. While there is strong intent on the part of these organisations to scale GenAI for external applications, most of them are in the process of getting their AI governance strategies right.

Broadly, two key considerations are impacting organisations’ deployments of GenAI. The first factor is return on investment—organisations today realise that one model does not fit all use cases. In fact, trying to retrofit large models available in the market to specific use cases can be very expensive. Therefore, picking the right model for the right use case has become critical. The second factor is the risks associated with GenAI—such as hallucination, drift, profanity, hate profiling, bias, lack of transparency, etc. Organisations are being mindful of implementing the right guardrails around GenAI and ensuring that their GenAI model is responsibly tailored to deliver the intended use case in line with the business objective.

What are the key challenges faced by Indian BFSI firms in implementing AI and GenAI strategies? How do they avoid issues like irrelevant or wrong answers or incorrect AI understanding of data?

Siddhesh: The primary challenge organisations are facing is how they can integrate GenAI with the workflow across their business. It is not just about identifying and scaling use cases, but ensuring that they have an organisation-wide strategy for GenAI and they can effectively integrate this technology across levels and functions. Another challenge organisations need to tackle is data complexity—how they can establish a robust data foundation to ensure data quality and provide a single source of truth across the business. This is especially important considering data complexity grows multifold as data is spread across hybrid and multi-cloud environments.

Most importantly, organisations must build the right AI governance capabilities—ensuring that their models adhere to the principles of explainability, fairness, robustness, transparency and privacy. AI governance is especially useful to help organisations avoid issues like hallucination, drift and bias among others. It provides robust processes to make sure the right training data is used, clear boundaries are defined for output consistency in line with the business objective, as well as rigourous evaluation and refinement of the AI model, as the organisation’s data evolves and ages. For example, using watsonx.governance, BFSI firms can monitor their customer service chats for red flags regarding toxicity, personal information leaks, or off-topic conversations and alert the concerned team when responses are outside of social norms.

How can BFSI firms approach and implement responsible AI in their organisations? What are the guardrails they should have in place as they use AI to drive business outcomes?

Siddhesh: There are three key areas that every organisation should focus on to build a proper AI governance solution. First, have a robust AI risk management framework—businesses need to use a combination of tools, practices and principles to identify, mitigate and address the potential risks associated with AI technologies. This includes data risks, model risks, operational risks, and ethical and legal risks. 

Second, ensure compliance with the evolving regulatory requirements, especially considering that the importance placed by regulators on responsible AI is now immense. Third, lifecycle governance is crucial—BFSI firms will likely be working with multiple AI models for different use cases in parallel. Therefore, continuously monitoring AI models (MLOps/LLMOps) to make sure they meet regulatory standards for model performance, as well as ascertaining there is no leakage of any customer PII becomes paramount.

How is IBM helping BFSI enterprises in India in their AI governance journeys? Can you provide some examples?
  
Siddhesh: In the BFSI sector, IBM has been helping organisations in their AI and GenAI journeys in four main areas—customer experience, operational efficiency, risk and compliance, as well as tech modernisation. With IBM watsonx, an integrated AI and data platform, IBM is helping BFSI firms scale and accelerate the impact of AI with trusted data across their businesses. 

For example, IBM has collaborated with one of India’s leading national banks to implement an AI governance framework. This framework effectively addresses bias prevention, ensures model explainability, and helps them meet stringent regulatory requirements. By managing the entire lifecycle of AI models—from robust data handling to deployment and ongoing monitoring—this engagement has significantly bolstered the bank's ability to deploy AI responsibly. Another example of IBM’s efforts is the recent MoU signed with the Department of Science and Technology, Government of Gujarat to establish and promote an AI Cluster powered by IBM’s watsonx to foster innovation and collaboration among financial institutions in Gujarat International Finance Tec (GIFT) City.

Further, watsonx.governance, one of the three core components of the watsonx platform, enables robust risk mitigation and bias detection, and enhanced transparency. This unified platform is designed to help financial institutions stay ahead in an evolving regulatory landscape, strengthen their data privacy controls and foster stakeholder trust in their AI-driven processes.