At a time when AI is moving from chatbots to doing real tasks for us, payments and commerce are also starting to change. We are no longer just talking about faster checkouts or smoother UPI flows. The bigger shift is about AI systems that can search, compare, decide and even pay on our behalf. That idea is being called ‘agentic commerce’. To understand what it really means and how close it is to becoming mainstream, I spoke to Nitendra Rajput, Head of AI Garage at Mastercard. In this conversation, he explains how agentic commerce goes beyond payments, the challenges behind it, and why he believes India could adopt it sooner than we think.
Below is the full interaction.
What does agentic commerce mean?
Nitendra Rajput: So, 30 years ago, if I wanted to buy something, I would go 3 kilometres, and if I did not find what I wanted within that 3 km, I would have to settle for a product available in my vicinity. If there might have been a better product at a lower price 7 km from me, I would not even have known of it. That producer who has produced that product would not know of a needy customer.
When e-commerce emerged, discoverability increased. So a consumer could get what they really wanted. A merchant could actually sell what they produced. Now we have so many choices. Today, if I want to buy something, I as an individual cannot figure out what all exists. So now we have agentic commerce. Agents on my behalf can swipe the entire web and find the most relevant item for me.
So that increases the matchmaking, which means the producer can now really find the specific niche customer that they want, and the customer can find a very specific product that they want without our need to be so sophisticated that we can figure out what all exists — from the most secure merchant with the best loyalty, any points that will give it to me with the best offers. So all of that can be optimised with agentic commerce.
Also read: Pichai and Hassabis: India uniquely positioned for AI leadership
For someone new to the term, what does ‘agentic commerce’ actually change for the user compared to UPI or saved cards today?
Nitendra Rajput: So agentic commerce lets you do commerce, which means commerce is not just payment, as you said. Whether it is UPI or saved cards, that’s the payments part of it. Before you come to payments, you need to figure out what needs to be bought.
There are so many options. What is most relevant for me? For that, I need to know the history of the user. So there is an agent that is figuring out what all exists. There’s another agent that is figuring out, oh okay, this person is a brand-conscious person versus this person is a value-for-money type of person, and so on. So based on my preferences, it can figure out what’s the most relevant thing for me.
A third agent will figure out that this thing is available at 16 different merchants. Which merchant looks the safest? Because these days a merchant can set up a shop, sell things for 80% discount, take all their card details, and vanish from the market in two days. So, is it a safe merchant? So it will identify if it is a safe merchant.
A fourth agent will identify that you have the product which is relevant for you and this is the safe merchant. Which of your payment options that you said is going to be the most beneficial for you from your loyalty, from your points, from a rewards perspective?
So this is what all these different agents are doing, and they are now giving you a very efficient commerce experience rather than the end payment experience that you’re talking about.
What was the toughest challenge in letting an AI agent complete a secure payment end-to-end?
Nitendra Rajput: I think the challenge is in the stitching of things together. So far, we’ve had situations wherein we all optimise on one thing. An e-commerce company could create a very good website for a person. The payment companies can create a very good mechanism to do payments, tokenisation, and so on and so forth.
Stitching the entire thing together is what makes the whole experience, because as a user, you are not looking at, okay, this e-commerce addressed this, this payment company did that, and the bank did that, someone else did the security stuff. I, as a user, just want that particular product.
So stitching all of that together, making them talk to each other in this standard language, is what was perhaps the most challenging.
With banks, aggregators and merchants involved, how hard is it to make AI-led payments work smoothly across everyone?
Nitendra Rajput: It is an ecosystem play. So the hardness — I mean I won’t say it is hard — but in some sense it is important for all of us, and that’s where events like these are useful, because all of us can get to know what others are doing and then these use cases evolve.
So a lot of discussion has to happen. A lot of design workshops have to happen. A lot of innovation forums have to be created to ensure that I can use what you are producing, you can use what I am producing, and we can together give a great experience to the end users.
If AI starts paying on our behalf, how do you make sure users stay in control and feel secure?
Nitendra Rajput: I mean, it sounds scary, but honestly, payments have been happening on our behalf as well. When we do subscriptions, we have monthly subscriptions for a lot of things.
Like you may have an OTT account and so on. So every month, some amount gets deducted from your card already, and we are fine with it. Now we don’t worry about it because we’ve gotten used to it.
It is just that those are already periodic subscriptions, which I know they’ll do at a specific point in time. Now they’ll become aperiodic because now you’ll do things more often. But machines have actually been making those transactions on our behalf already.
Do you see India adopting agentic commerce quickly, or is this still a long-term play?
Nitendra Rajput: I’m bullish about the opportunity that it will provide with respect to what we started from — matchmaking. It will make this whole economy more efficient, and who doesn’t want efficiency? So, I’m very bullish that India will adopt it very soon.
Unfortunately, these guesses keep failing us because these days, the technology adoption in the masses has improved significantly. Earlier, it used to take, let’s say, 10 years for something to become mainstream from the innovation side of it. Now it takes like a few months. So from that point of view, I think it will be very soon rather than much later.
Also read: IndiaAI Impact Summit 2026: India’s move from buying to making semiconductors