Indian fintech company Razorpay is reimagining its platform for an AI-driven future. Chief Product Officer Khilan Haria says the firm is effectively rebuilding itself from the ground up, with AI at its foundation.
In an exclusive interview with AIM on the sidelines of FTX26, Haria discusses how the IPO-bound company has spent the past year rethinking its products for a world frequently shaped by conversational interfaces, AI agents and automated workflows.
Towards Conversational Fintech
At the event, Razorpay launched Agentic Experience Platform, an AI layer that simplifies onboarding, payment integrations and operational workflows for merchants.
The platform includes three features: agentic onboarding, agentic dashboard, and agentic integration.
Agentic onboarding allows businesses to submit details such as PAN and website information, while the platform automatically completes identity verification and compliance checks through government systems. This reduces onboarding time from about 30-45 minutes to roughly five minutes.
The agentic dashboard allows businesses to perform operational tasks using natural language queries. For instance, merchants can upload bank statements and ask the system to match them with Razorpay settlements.
Agentic integration allows developers and partners to integrate Razorpay’s payment infrastructure using AI coding tools such as Claude Code and platforms like Replit.
Instead of requiring businesses to navigate menus or learn product-specific workflows, Haria says the goal is to let users simply describe what they want. “Every interaction and every touchpoint should look very natural, how they are used to conversing with a friend or a confidant,” he adds.
Working with Multiple Models and Agent Studio
Razorpay is working closely with AI leaders such as OpenAI and Anthropic as it builds its AI stack.
“We partner with them at a very deep level to continuously discover customer problems and solve them,” Haria says, adding that each company brings distinct strengths. “Optimisation continues to happen heavily on whichever models allow us to solve for accuracy and latency.”
For instance, using the Claude Agent SDK, the company has also introduced Agent Studio. This platform allows businesses to deploy AI agents that can manage payment-related tasks such as recovering failed transactions, managing disputes and forecasting cash flow.
The idea, Haria says, is that Razorpay cannot build every use case internally. Instead, it is creating a marketplace where developers can build agents using its data pipelines and infrastructure.
“If we create the infrastructure of building these agents and expose the context they need, anyone can build solutions,” he explains.
By opening up these capabilities—some proprietary and some publicly available—the company hopes customers will be able to build their own agents rather than depend entirely on Razorpay.
Haria shares that for internal company workloads, developers are already using tools such as Claude Code, Codex, Antigravity and Cursor. Razorpay is also embracing vibe coding platforms such as Replit and Emergent, so developers can build integrations more quickly.
Haria says Razorpay is taking a flexible approach to its tech stack, supporting multiple standards rather than betting on just one.
“It’s a mix of everything—we are infrastructure providers,” he says, adding that the company was among the first in India to adopt MCP as it emerged as an open standard. He believes APIs, MCPs and even CLI tools will coexist, as developers increasingly rely on different environments and workflows.
Agentic Commerce
Looking ahead, Haria says that Razorpay is preparing for the rise of agentic commerce, where AI assistants make purchases or complete transactions on behalf of users.
He believes India’s consumer behaviour may accelerate adoption. “Fundamentally, many of us in India are ‘do-it-for-me’ users. If we can delegate a commerce activity to someone, we will.”
Razorpay is collaborating with the National Payments Corporation of India to support this transition through initiatives like Reserve Pay and UPI Circle, which enable delegated and automated payments.
The company demonstrated early examples of conversational commerce integrations with platforms such as Zomato and Swiggy.
Haria also pointed out that India’s mobile-first ecosystem could help the country leapfrog global markets again, much like it did with mobile payments.
The next phase of UPI, he suggests, will likely be driven by biometric authentication and new payment use cases.
“Biometric authentication is becoming real, and over the next few years, most of us will possibly keep entering biometrics instead of PINs,” he says, adding that trends such as credit cards on UPI and agentic commerce are expected to further accelerate adoption.
Razorpay is also exploring voice-based payment. Through integrations with voice AI platforms such as Gnani.ai, automated agents can assist in collecting payments.
Razorpay’s strategy reflects a broader shift across technology companies as AI becomes the primary interface between users and software. When asked to describe the company after its latest product event, Haria summed it up in one line: “Razorpay—reimagined and ready to win in the AI era.”
Your reaction
Discussion
What's your take on this story?
No comments yet — be the first to share your view.