UPI already moved India past cash and past cards. The next shift is smaller but stranger: instead of a person opening an app and tapping "pay," an AI agent may do it for them.
NPCI is building the infrastructure for that. It's called the Unified Agent Protocol (UAP), and if it launches the way it's currently described, it changes who — or what — is actually completing the purchase on your store.
What the Unified Agent Protocol actually is
UAP is not a new payment rail. UPI stays UPI. What UAP adds is a way to register, verify and authorise an AI agent to transact on a person's behalf, using UPI Circle's existing delegated-payments framework as its foundation rather than building something from scratch.
In practice, that means:
- A shopper grants an AI assistant permission to pay for specific things, up to a spending limit they set.
- The agent can complete a payment without asking for approval on every single transaction — the point is to remove the friction, not just move it.
- Every payment the agent makes is still bound by consent controls the shopper defined upfront. An agent can't decide to spend outside its limit, and it can't pay a merchant it wasn't authorised for.
The first use cases being discussed are the boring, repetitive ones: topping up recharges, paying utility bills before the due date, renewing subscriptions, reordering groceries. Not because that's the ceiling — because it's the easiest place to prove the model works before anyone trusts it with anything bigger.
One thing worth being precise about: UAP still needs a regulatory go-ahead from the Reserve Bank of India before it launches. This is real infrastructure being actively built, not a shipped product yet.
Why this matters more in India than almost anywhere else
Agentic checkout is being built by OpenAI and Google too — ChatGPT's Instant Checkout runs on the Agentic Commerce Protocol, and Google has its own Universal Commerce Protocol and AP2. Both depend on card-network rails and tokenised payment credentials built for markets where cards are the default.
India isn't a card-first market. UPI already handles a scale of everyday, low-value, real-time transactions that most countries don't have an equivalent for. If NPCI gets agentic payments right, an AI assistant paying for your customer's groceries on UPI could be simpler and faster to build than the same flow on a card network anywhere else. That's a real, structural advantage — and it means agent-initiated buying may normalise here faster than in the US or Europe, not slower.
What this means for your store, today
None of this works if the AI agent can't tell what you sell, at what price, or whether it's in stock.
An agent authorised to spend ₹500 on a reorder still has to know your product exists, know it costs ₹499 and not ₹599, know it's actually in stock right now, and have a way to actually check out — before it can spend a single rupee with you instead of a competitor.
That's true regardless of which protocol eventually wins in India — UAP, ACP, UCP, or something that doesn't exist yet. The requirement underneath all of them is the same: your store has to be legible to something that isn't a person. Product names, prices and stock status that are accurate and machine-readable, not just pretty on a webpage.
Two free ways to check where you stand:
- Check whether AI recommends your brand — see if ChatGPT, Gemini and Claude are already naming you when shoppers ask what to buy.
- Check if AI agents can actually read and buy from your site — the same technical readiness an agent needs before it can complete a UPI payment on your behalf.
Being recommended and being purchasable are two different problems. UAP is a preview of how fast the second one is about to matter.
StockedBy checks whether ChatGPT, Gemini and Claude recommend your brand, and whether your store is ready for AI agents to buy from it — free, for brands in India, UAE and Saudi Arabia.