Your bank account could one day become an AI agent: Perfios Group CEO Nitin Chugh

What if your bank account did more than hold your money? What if it could understand your financial position, make recommendations and even stop you from making a bad financial decision?

That may not be as far-fetched as it sounds.

At the Global Fintech Fest 2026 in Mumbai, Nitin Chugh, MD & Group CEO of Perfios, said the evolution of agentic AI could eventually turn the bank account itself into an AI agent — one that can act on behalf of the customer.


“Theoretically, your bank account itself could be an agent,” Chugh said, adding that at a much more advanced stage of AI, such an agent could potentially prevent a customer from making a poor financial decision.

Artificial intelligence (AI) in banking is moving beyond chatbots, copilots and tools that simply help employees work faster. The next phase could involve AI agents that coordinate with one another, make decisions within defined guardrails and execute parts of financial workflows.

Also Read: If attackers are using AI, companies will have to use AI to defend themselves: Razorpay CEO

Perfios is already positioning its technology around this shift. At GFF 2026, the B2B Techfin company showcased its next-generation Agentic AI Operating System, which it says is designed to help financial institutions use real-world, everyday data to assess credit.

For Chugh, however, the more transformative opportunity lies in what happens when these capabilities start working together.

Consider a home loan.

Today, getting a loan approved can involve multiple steps — customer onboarding, document collection and verification, credit assessment, decisioning, sanction and finally disbursal. Chugh said an agentic workflow could potentially bring several of these processes together, with 15-20 agents communicating with each other.

A process that could take two days, he said, could eventually be completed in five to 10 minutes.

Under the model Chugh described, the relationship manager could continue to handle the customer conversation while AI agents work in the background on operational and decisioning tasks.

That is where agentic AI starts looking different from simply putting an AI layer on top of existing banking software.

Banks have already been using AI and machine learning for areas such as fraud detection, customer onboarding, document analysis and early-warning signals. But agentic AI could connect these individual capabilities and allow systems to take action across an entire workflow.

Also Read: GFF 2026: BharatPe launches agentic AI assistant for merchants, connects to over 60 live systems

Chugh said banks could therefore find themselves taking two very different approaches to AI over the next few years. Some could use the technology primarily to drive efficiency and reduce costs. Others could redesign their operating models around AI.

But giving AI agents more autonomy also brings a new problem: what happens when the agent goes rogue?

Chugh said agents are different from traditional AI models because they can learn, adapt and behave differently depending on the situation. The guardrails built around conventional large language models may therefore not be enough when AI systems are making decisions and acting independently.

The controls, he said, will need to evolve faster than the agents themselves. It also changes the cybersecurity equation.

Banks will not only have to protect themselves against human attackers. They could increasingly have to defend against AI-powered systems operating at machine speed.

Chugh said that if sophisticated AI systems are being used by attackers, financial institutions will need an even more advanced layer of AI to defend themselves — effectively an AI-versus-AI security race.