Fintechs push AI, new protocol to simplify backend tools for merchants

Cashfree and Razorpay use AI agents and a new protocol to simplify backend systems and automate payments, easing merchant operations across fintech platforms

Business-to-Business (B2B) fintech companies like Cashfree Payments and Razorpay are making it easier for businesses (merchants) to manage complicated backend systems. They are doing this with the help of artificial intelligence (AI)-powered agents and a tool called “Model Context Protocol” (MCP).

 

The MCP is like a “universal connector” — in other words like a USB-C port — that allows different software systems to work together easily. This enables AI agents and assistants to interface directly with core application programming interfaces (APIs), streamlining integration for merchants of all sizes.

 

Without it, businesses would require manual and complex integration of compatible AI systems with every element of their fintech infrastructure, which includes applications such as payments, verification, payouts, among others.

 

“The idea was to integrate large language models (LLMs) with our own tools and APIs, and make the experience smarter. Our core APIs is what the LLM would need for it to become useful,” said Nitin Pulyani, senior vice-president (SVP), Product, Cashfree Payments.

 

Additionally, for the AI agents of merchants and businesses alike, their fintech experience is now tailored to be conversational and standardised.

 

“The MCP clients — (AI) assistants and agents — will evolve where voice will become more natural. We will have to keep enhancing as the protocol evolves for new use cases that the industry will start using,” said Khilan Haria, chief product officer, Razorpay.

 

Use cases

 

In its current avatar, an MCP is designed to handle multiple use cases, with payments being the dominant one.

 

Merchants, especially smaller ones, had to rely on a separate communication mediums, such as WhatsApp or short messages, to communicate with their customers to generate payment links, confirm transactions, or initiate refunds.

 

With AI, MCP and payment APIs, the entire process can be automated after being initiated in a natural language like English. 

 

“The model understands the merchant’s menu and the rate card. When a customer orders a particular item, it will understand that a payment is supposed to be made and search for payment APIs in the background since it is integrated,” Pulyani explained.

 

He added that these APIs will then generate a payment link and once the payment is made, it will look for the status of the transaction.

 

Haria added that one of the emerging use cases is, AI agents will interact with the company modules to get answers to queries instantly. Operations teams can expose the fintech’s dashboard inside their own agents and assistants to enable procedural ease.

 

“Our hypothesis is this will be very big. In the next 6-12 months, this will become a way of living. We are expecting massive scale in AI usage henceforth,” he explained.

 

Possible constraints?

 

Product heads explained that despite the scale with which AI evolves, they don’t expect any constraints on these newer systems yet.

 

This is primarily because an MCP is an intermediary between multiple moving parts in the form of agents and core APIs. The fintech APIs, which form the core product for companies, continue to scale as customer base and capabilities ramp up.

 

“The capacity (to handle) is not dependent on the MCP since it can be deployed on a machine. It’s the same that is required for any service deployment, and is instead dependent on the capacity of the LLM in use,” Pulyani said. 

Making it simple 

Fintech companies doing this with help of AI-powered agents and tool called “Model Context Protocol”

Fintech experience now tailored to be conversational and standardised

With AI, MCP, and payment APIs, entire process can be automated 

Product heads don’t expect constraints on newer systems yet

Fintech APIs  continue to scale as customer base