First, What Is Agentic Commerce?
At its core, agentic commerce is about delegating the act of purchasing to an artificial intelligence 'agent'. Instead of you manually browsing an app, adding items to a cart, and entering payment details, you simply give instructions to an AI. For example,
you could tell a chatbot, "Order my usual weekly groceries from BigBasket," or "Book the earliest flight to Mumbai for tomorrow morning under ₹5,000." The AI agent would then handle the entire process—finding the items, placing the order, and completing the payment—autonomously. The shift moves commerce from being transactional, where you perform every step, to being conversational and intent-driven, where you just state your goal and the AI executes it.
The Power Couple: UPI and AI
This futuristic shopping experience is made possible by combining AI with the massive, real-time payment infrastructure of UPI. India's UPI network is one of the largest in the world, processing billions of transactions monthly. The National Payments Corporation of India (NPCI) is now developing a framework, reportedly called the Unified Agent Protocol (UAP), to allow these AI agents to securely make payments on a user's behalf. This isn't just a theoretical idea; pilot programs are already underway with companies like OpenAI, Razorpay, and major Indian merchants such as BigBasket, Zomato, and Swiggy. The goal is to transform UPI from a system operated by humans into a programmable network where AI can execute transactions based on pre-approved rules.
How It Works: Control and Convenience
The system is being designed with security and user control as top priorities. It's expected to build upon two existing UPI features: UPI Circle and Reserve Pay. UPI Circle allows a primary user to delegate payment authority to another user, which in this case would be a verified AI agent. Reserve Pay lets you block a certain amount of funds in your account for a specific merchant or purpose, creating a pre-approved spending limit. So, you could authorize your AI agent to spend up to ₹2,000 per week on groceries. The agent can then make multiple small purchases within that limit without asking for your PIN every single time. You retain full control, with the ability to set rules, monitor transactions, and revoke access instantly.
The Impact on Online Shopping
For e-commerce, this could dramatically reduce friction and 'cart abandonment'. The entire journey from product discovery to payment could happen within a single chat window, without redirects or complicated checkout pages. This conversational commerce model makes shopping more natural and efficient. For example, an AI agent could monitor for a price drop on a product you want and automatically purchase it when it hits your target price. This changes the focus for businesses from optimizing clicks to optimizing for customer outcomes, making sure they can be trusted by a user's AI.
Revolutionizing Offline and Hyperlocal Retail
The impact isn't limited to large online platforms. Agentic commerce could be a game-changer for offline retail and India's Open Network for Digital Commerce (ONDC). Imagine telling your AI, "Find the nearest kirana store that delivers and order two litres of milk and a loaf of bread." The agent could interact with ONDC-listed local stores to find the best option and complete the purchase. This could empower millions of small and medium-sized businesses by giving them access to AI-driven demand without needing to build their own complex apps. It levels the playing field, allowing even the smallest neighbourhood shop to compete in a conversational marketplace.
Challenges on the Road Ahead
While the potential is enormous, there are significant hurdles to overcome. Establishing a clear liability framework is crucial: who is responsible if an AI agent makes a mistake or an unauthorized purchase? Defining responsibility among the user, the bank, the AI provider, and the merchant is a complex challenge that NPCI is working to address. Furthermore, ensuring robust security to prevent new types of fraud, managing data privacy under regulations like the DPDPA, and building user trust are all critical for widespread adoption. The technology needs to be not only seamless but also demonstrably safe and reliable.
