What is agentic commerce?
It is shopping where an AI agent does some of the work for the buyer. Today a buyer asks an AI app for “a cotton bedsheet under ₹1,500” and then clicks through to a store. With an agent, the buyer can say “reorder my usual tea” and the agent finds it, fills the cart and pays. The buyer sets the limits. The agent acts inside them.
What is actually live in India?
Payments are moving first. In February 2026, Razorpay and NPCI announced agentic payments on Claude at the India AI Impact Summit. Users could order from Zomato, Swiggy and Zepto inside a chat. Razorpay said the platform was in a pilot phase with a small group of users. It runs on UPI Reserve Pay, where a user approves one spending limit for a merchant. The user can also withdraw that consent at any time. Note what this pilot is and is not. The agent orders from a few large apps, inside a limit the user approved once. It is not yet an open system where any agent can buy from any small store.
What about shopping inside ChatGPT?
Discovery is further along than buying. OpenAI’s merchant page says users can discover and compare products in ChatGPT, but purchases are completed on the merchant’s own website or app. The same page says shopping is live in the U.S. and will expand to more regions over time. India is not named yet. So for an Indian store, the near-term win is being found, not being bought from by a bot.
What stays the same and what changes?
| Part of a sale | Today | With agents |
|---|---|---|
| Finding products | Search, social and ads | AI apps compare pages and prices for the buyer |
| Choosing | Buyer reads the page | Agent reads the page, then the buyer approves |
| Paying | UPI, card or COD at checkout | UPI within a limit the buyer set |
| After the sale | Your returns and support | Still your returns and support |
What should a small store do now?
- Write product titles that say exactly what the item is, with size, material and colour.
- Show one clear price, delivery charge and delivery time on every product page.
- Keep return, refund and cancellation policies short and easy to find.
- Check your robots.txt does not block AI search crawlers you want to reach.
- Keep stock counts accurate, so an agent never buys something you do not have.
What are the risks for sellers?
The main one is a wrong order with no human reading it first. An agent may pick the wrong size if your page is vague. Disputes may also take longer while the rules settle. Keep your policies clear and your order notifications on. Watch for unusual repeat orders, too. None of this needs new software today. It needs pages that a careful machine and a careful person read the same way.
Getting ready on your own store
On The Storemate, product pages already carry structured product data, and you can edit your store’s robots.txt under SEO. You can also connect Claude or ChatGPT to your own admin with read-only access. That gives you a feel for how AI apps read your store before buyers’ agents arrive.