India’s financial sector eyes ‘productivity revolution’ through Agentic AI

The initial wave of Artificial Intelligence (AI) for the Indian financial services industry was defined by experimentation is moving away from simple automation toward an agentic model where artificial intelligence acts as a digital colleague to enhance productivity rather than just a tech tool.

At a recent industry session organised by Salesforce in Delhi, top executives from India’s leading financial institutions and technology firms gathered to deconstruct the gap between AI’s “cool factor” and real-world business value. The consensus was clear – the next phase of growth will be defined not by the technology but the integration of autonomous agents into the operating fabric of the enterprise.

Sharing his personal shift from traditional workflows to using AI agents like Claude and Gemini for complex tasks, Bajaj said: “I don’t look at AI as a technology revolution. It is a productivity revolution that is happening. People who know how to use AI are going to be 10x more productive or 10x faster than the people who can’t use AI.”

“A lot of AI pilots are failing or not being able to scale because we do these pilots more as a proof of technology rather than a proof of business value. When you talk about scaling, you go to your leadership and the inevitable question is, ‘What’s the ROI?’ and there is no clear answer to it,” he said.

This lack of clarity is often compounded by the noise of fragmented corporate data. Nikhil Garg, Chief Operating Officer at Glimmer Technologies, warned that many companies lose sight of the core problem.

“One of the major challenges in the AI projects overall is the scope of the business problem. Everybody at this moment is experimenting a lot on AI and that's where the scope of what business problem we are going to solve probably goes for a toss,” he said.

“One of the big things that we found out is when you’re trying to scale, how do you actually put evaluations for AI? AI actually confidently hallucinates, and I don’t think there is a clear mechanism of actually evaluating what the response of an AI should be or where it is going wrong at scale,” he said.

To counter this, technology providers are embedding trust layers directly into the software. Sonali Kalyanikar, AVP at Salesforce, explained how security and context are the antidotes to AI errors.

“We understand what it takes to make sure that our customers get a trustworthy platform which does not happily hallucinate and gives context. It’s not only the AI platform, but the whole context that is through our CRM and the data 360 layer that helps our customers get outcomes,” she said.

“We’re not talking about AI taking over humanity. Agents are there to augment humans so that they can provide better customer experience. We’re far away from that world where software and people were just used to do static tasks, and we’re now talking about agents delivering outcomes.”

“Today 80-90% of our tickets are being resolved by our own agentic enterprise layer. Whatever ticket each of our customers raises today, the interface that you’ll see is a direct prompt and it collates all the problem statements and challenges worldwide,” he said.

Note to Readers: This article was produced as part of Salesforce's Strategic Leadership Dialogue. Mint is a media partner.