Facing Roadblocks: The Checkout Question for AI Commerce in India

Here's the honest answer first: AI commerce in India is not stuck because Indian AI is behind. It's stuck because Indian payments were deliberately built to need a human in the loop, and no AI agent, however smart, gets to skip that step yet.

AI commerce, sometimes called agentic commerce, means an AI assistant does the shopping for you. You tell ChatGPT, Perplexity or a WhatsApp bot 'I need running shoes under Rs 3,000, size 9', and the AI searches, compares, picks the best option, adds it to cart, and in the fullest version, pays for it too. Browsing, comparing and cart-building are already working well, including in India. The trouble starts at the very last step: the actual payment. That's the 'checkout question' Business Standard flagged, and it's a real one, because payment is the one part of the flow where Indian regulation intentionally does not let automation run free.

In the US, OpenAI's Instant Checkout and similar agentic-payment pilots work because cards are tokenised and stored, and a single authorisation can often clear a transaction without a fresh human step every time. India works differently. Two-thirds of India's digital retail payments run on UPI, and UPI is designed around a human physically entering their UPI PIN on their own device for every transaction. There is no 'stored PIN' an AI can use on your behalf, by design. Even card payments in India need a second factor almost every time. So the AI agent can do 95% of the job, search, compare, negotiate, build the cart, and then it hits a wall it cannot legally or technically cross alone.

The Reserve Bank of India has required Additional Factor Authentication, AFA, for card and UPI transactions since 2019, meaning a second, independent proof of identity, an OTP, a UPI PIN, a biometric, on top of the basic transaction details. RBI also tightened card-on-file tokenisation rules from October 2022, which stopped merchants and platforms from storing raw card numbers at all. Both rules exist for genuinely good reasons, they've cut card fraud sharply. But they also mean no AI agent, no matter how well trained, can complete a payment on your behalf without you personally authenticating it in that moment. This is the single biggest structural difference between 'AI commerce in the US' and 'AI commerce in India', and most global coverage of agentic checkout glosses over it.

Even outside metros, Cash on Delivery still accounts for a meaningful chunk of ecommerce orders in India, industry estimates put it anywhere from 30% in top cities to well over 50% in tier 2 and tier 3 towns for categories like fashion, home and general merchandise. COD orders need a phone confirmation or an OTP-based order verification to cut down fake orders and returns, which is again a human-confirmation step an AI cannot do on its own. So even if an AI agent nails the product recommendation and the cart, a large share of Indian buyers still expect a real conversation, a real confirmation call or message, before the order is locked. That's not a technology gap, it's a trust and fraud-control gap, and it needs to be designed for, not automated away.

Visa announced its Trusted Agent Protocol and Mastercard its Agent Pay in 2025, both trying to create a standard way for AI agents to be recognised as trusted intermediaries at checkout, similar to how Google's Agent Payments Protocol, AP2, is trying to standardise agent-to-merchant payments globally. These are genuinely important developments and worth watching. But they are being built card-first, for markets where cards dominate. India's payment stack is UPI-first, RBI-regulated, and built around real-time human authentication by design. Until these protocols are adapted specifically for UPI's AFA requirements, and RBI signs off on any exception, full autonomous AI checkout in India is going to lag the US and Europe by a fair margin, regardless of how good the underlying AI gets.

The AI can do the thinking. In India, the human still has to do the tapping.

This is good news, not bad news, for most Indian small and medium businesses, because it turns out the workaround is already how Indian consumers like to shop. Instead of chasing 'fully autonomous AI checkout', the realistic and honestly more trustworthy model is: AI does everything up to the payment, then hands off a clean, pre-filled payment link or UPI QR that the human confirms in one tap. This is exactly the shape of conversational commerce that's been quietly working on WhatsApp for years, an AI agent chats with the customer, understands what they want, shows them a catalog, builds their cart, answers objections, and then sends a payment link, so checkout becomes a two-second confirmation instead of a multi-step form. No regulation is broken, no trust is lost, and the customer still feels like a human, or a very smart assistant, sold to them, not a black box.

You don't need to wait for Visa, Mastercard or RBI to sort out agentic protocols for India. You can build most of this workflow today, using tools that already respect Indian payment rules. A practical setup looks like this: an AI agent trained on your product catalog and FAQs handles the conversation, a connected commerce layer shows your catalog with prices and images inside the chat, the AI builds the cart based on what the customer says, and a payment link or UPI-based checkout is generated automatically the moment the cart is confirmed, with the customer completing their own PIN entry. For COD orders, the same AI agent can trigger an automated confirmation message instead of a manual call, cutting down on RTO, return to origin, which is one of the biggest silent cost-drains for D2C brands in India, often quoted between 15% and 30% of COD orders depending on category.

This exact gap, AI doing the conversation and the cart, human doing the one-tap payment, is what ODIV's ai-workflow-automation service is built around. Our team designs and builds the full backend for you: an AI agent trained on your business that talks to customers on WhatsApp, Instagram or your website, a workflow that moves them from question to catalog to cart automatically, and a checkout handoff using WhatsApp commerce, payment links or UPI QR that respects every RBI rule while still feeling instant to the customer. If you already use, or plan to use, ODIV Engage for your WhatsApp catalog, cart and payments, we connect the AI automation directly into it, so leads move from a first 'hi' to a confirmed, paid order without your team manually chasing anyone. You get the speed of AI commerce and the compliance and trust of the Indian payment system, without gambling on protocols that aren't built for UPI yet. If this sounds like what your business needs, just start a chat with us on WhatsApp and we'll map out the workflow together, no pressure, just a practical look at what's possible for your catalog and your customers.

AI commerce in India will keep getting more capable every quarter, that part is not in doubt. But the checkout question isn't really a technology question anymore, it's a design and trust question, and the businesses that solve it first, with a smart AI-plus-human handoff, are the ones that will win the next few years of Indian ecommerce, not the ones waiting for full autonomy to arrive.