Gnani.ai, Razorpay Launch Agentic AI Platform That Completes Payments During Live Calls

Bengaluru-based voice AI startup Gnani.ai has partnered with Razorpay to launch what the companies call the industry’s “first agentic AI collections platform capable of completing payment transactions during live customer calls”.

The platform integrates Gnani.ai’s enterprise voice AI system, which processes over 10 million calls daily, with Razorpay’s Model Context Protocol (MCP) server, an AI-native payments execution layer that enables AI agents to initiate, authorise, and complete transactions in real time.

The Agentic Collections AI platform is available for enterprise deployment. Financial institutions can request demonstrations and implementation proposals directly from Gnani.ai.

The companies said the integration addresses a long-standing gap in automated collections, where voice AI systems could remind customers about dues but could not close transactions within the same interaction.

“For the first time, enterprises can automate collections at scale without sacrificing conversion,” said Ganesh Gopalan, Founder and CEO of Gnani.ai. “This integration delivers what voice AI has promised but never achieved: the ability to complete transactions, not just initiate conversations. 

Through the integration, Gnani’s AI agents connect directly to Razorpay’s MCP Server, allowing them to generate payment links, initiate UPI collect requests, monitor payment status, and confirm settlement within the same call.

The AI assesses customer intent, determines suitable payment instruments, and executes the transaction flow without requiring human intervention for routine merchant-side actions.

The platform currently supports standalone UPI payments for one-time transactions and UPI Single-Block-and-Multiple-Debits (SBMD), also known as Reserve Pay, mandate-based payments for recurring obligations. Support for cards and wallets is planned.

Khilan Haria, Chief Product Officer at Razorpay, said the integration enables AI systems to move beyond conversational engagement. “Agentic AI only becomes truly valuable when it can act, not just converse. With Razorpay’s Agentic Payments infrastructure powering this integration with Gnani.ai, we’re giving AI systems a secure, real-time pathway to complete payments inside live customer conversations. This moves collections from ‘follow-up’ to ‘finished’ in a single interaction.”

Razorpay’s MCP Server acts as an interface between Gnani’s agents and Razorpay’s payments infrastructure, allowing AI-driven commands to generate links, check status, and confirm settlement without relying on traditional dashboards.

Gnani.ai said its proprietary automatic speech recognition technology improves accuracy by 30% over competing systems and supports conversations in more than 38 languages, including Hindi, Bengali, Punjabi, Tamil, Telugu, Marathi, Gujarati, and Kannada. 

The platform verifies outstanding amounts, confirms payment intent, generates transaction-specific links, sends them via SMS or WhatsApp, and tracks payment status through polling and webhooks during the same interaction.

The companies stated that customer payment details do not pass through the voice AI system and are processed through Razorpay’s infrastructure. Payment links are generated with transaction-specific parameters to prevent reuse. UPI SBMD mandates require customer authorisation through registered UPI applications.

Gnani.ai said its platform maintains GDPR compliance and SOC 2 Type II certification.

Razorpay, in collaboration with the National Payments Corporation of India, has launched an agentic payments pilot on Anthropic’s Claude.

It lets users complete transactions, such as ordering food from Zomato and groceries from Swiggy or Zepto, directly within a conversational AI interface, without switching apps or entering payment details manually. 

Meanwhile, Gnani.ai  launched two new foundational voice AI models at the India AI Impact Summit 2026, including a five-billion-parameter voice-to-voice system and expanded speech-to-text and text-to-speech models under its Inya VoiceOS stack, as part of India’s sovereign AI push.