GFF: AI could soon turn bank accounts into agents

Imagine a scenario where your bank account becomes more than just a storage unit for your funds. Instead, it could evolve into an artificial intelligence (AI) agent capable of understanding your financial situation, offering recommendations, and even preventing you from making poor financial choices. This is not a far-fetched idea, but a potential reality according to Nitin Chugh, the Managing Director and Group Chief Executive Officer of Perfios.

At the Global Fintech Fest 2026 held in Mumbai, Chugh discussed the future of agentic AI in banking. He proposed that your bank account could potentially function as an AI agent, capable of executing tasks on behalf of customers. At a more advanced stage of AI development, such an agent could even intervene to stop a customer from making a bad financial decision.

AI in banking is progressing beyond traditional chatbots and copilots. The next frontier involves AI agents that collaborate with each other, make autonomous decisions within predefined boundaries, and carry out portions of financial workflows. Perfios is already developing its next-generation Agentic AI Operating System, aimed at enabling financial institutions to leverage everyday data for credit assessment.

However, the real game-changer lies in the integration of these capabilities. Consider the process of obtaining a home loan. Currently, it involves several steps, such as customer onboarding, document collection and verification, credit assessment, decision-making, sanction, and disbursement. Chugh envisions an agentic workflow where multiple processes could be orchestrated by several agents, potentially reducing the approval time from two days to just five to ten minutes.

This model represents a significant shift from merely adding an AI layer on top of existing banking software. Banks are already employing AI and machine learning in various areas such as fraud detection, customer onboarding, document analysis, and early-warning signals. However, agentic AI could connect these individual capabilities and allow systems to take action across an entire workflow.

As banks consider adopting agentic AI, they will likely face two distinct paths. Some may focus on using AI primarily to streamline operations and cut costs. Others may restructure their business models around AI. Regardless of the path chosen, giving AI agents more autonomy also introduces a new challenge: what happens when the agent deviates from its intended function?

Unlike traditional AI models, agentic AI can learn, adapt, and behave differently based on the situation. This autonomy may necessitate the evolution of control mechanisms beyond those employed for conventional large language models. Chugh emphasized the need for these controls to keep pace with the agents' learning capabilities. Moreover, the implications extend to cybersecurity, as banks may need to defend not only against human attackers but also against AI-powered systems that operate at machine speed.