Razorpay CEO Harshil Mathur on how agentic commerce will unfold in India

0:00 Today's episode of Zeroot is actually very special. I know you've been saying that every week, but this week is actually very special. Brady, do you want to tell us why it's a very special episode? >> Well, one aspect is that it's very, very forwardlooking. We're really looking at the future. So, what are we looking at? >> Okay, the other thing is where I'm recording from. I'm actually recording from not from the Ken office. I'm recording from another office in Kormagla and it's an office of a company that most of my listeners have heard of and this is of course Razer Pay and I'm going to begin by telling you why I'm here for almost a decade the most important object in Indian commerce was the checkout page and ready I'm sure this is true in China as well I don't know how checkout pages are there in China how are they in China are they like chaotic crammed they're trying to sell you 50 extra things sounds almost exactly like what it is over here. I feel like the China one is probably more chaotic from what I know. But in India, what happened was not only was this like so important, but then we had UPIU came along and what UPI did was it just made the act of payment feel so simple and so easy. You just had to put a fingerprint.

1:13 You just had to like put one chime and you're done. But you had to still do the payment. You still had to open the app. You still had to search. You still had to scroll. You still had to filter. you have to add to cart and every step along the line is where you know merchants and the sale would technically die and the reason why I'm here in the Razer pay office is because Razer Pay's bet is that this entire rigmarole and this choreography disappears at some point in time a customer wakes up and says I want to make halva tonight and halva is a very relevant example Brady do you know what halva is >> a sweet thing right >> it's a sweet thing you're right and so what happens is something all that all that the customer does is say that and then someone figures out the ingredients, bills the cart, checks what's in your pantry and just makes the conversation right inside without the app itself. And the company that processes $180 billion a year and sits underneath all of these other companies we've heard of like Zomato, Swiggy and Cred. This company calls this agentic commerce and this company's CEO thinks that it is India's next leap after UPI.

2:20 And so today's episode is not about all the various little things in it. We're just going to start by assuming that it's real. We're going to assume that in a few years an agent is going to do a lot of this buying. Then we're going to ask the question, what does that look like? What happens when the machine picks the restaurant? What happens to the restaurant when it goes and decides who delivers it? Who delivers it? When commerce moves to a chat window and something that just when you just ask for it and it happens. How do how does all of this work and who really wins and who's going to change in this future? And for this conversation, we have Harshel Matur, co-founder and CEO of Razer Pay, who has been talking about agentic commerce for such a long period of time to join us for this episode.

3:05 Welcome, Harsh. >> Thank you, Pravin. Thanks, Brady. Nice to talk to you guys. >> >> I just want to tell for our listeners that something very important which is that Razer Pit did not set any conditions on this conversation. They did not tell us what to talk about, what not to talk about. They didn't ask to see questions. They didn't want to steer it. So, we are just here to talk about all of the things that Agentic Commerce exists. The only thing we're not going to talk about today is about Razer Pay's business and its future projections and revenue simply because they've just filed for a DRHP. This news is public and immediately after filing for a DRP, a company goes through a silent period.

4:07 So, we can't talk about that. But despite that, we're going to be talking about agentic commerce a lot. So thank you Hershel. I really appreciate it. >> Yeah. No, great to talk to you guys and I'm very excited about agent commerce in general. So always happy to talk about it. >> Okay. So let's begin. I'm going to start with the first broad thing for our listeners without getting into all the details of the products that you have launched over the last 6 8 months which we will get to. I think one of the things that you had said in an earlier conversation was the chat is like a starting point and that AI should like finish the job. So it should not just recommend. It's not like so basically it's almost like you say it and it gets done. Can you like walk us what this going to look like? I of course we're not there yet today but let's say 2 years 3 years from now what does this look like? What is this thing from start to finish? How do you imagine it happening?

4:57 >> So think of today right like today when you want to do something new your gateway to internet starts from search which is where Google made made its big bucks that it starts with search. sometimes it starts with social or or when it's discovery so Instagram or something like that but a lot of it starts with search when you want to do something pointed hey to your example I want to make a halva it starts with search how to make a halva and then you figure out how to make it then you discover what you need to add to it and then you figure out an app through which you can order some of those things and so on and so forth increasingly we are seeing that that gateway to the internet is becoming LLMs or chats where India is the second largest user base for chat GPT in India. and I can say for myself personally that now when I have a question it starts with chat GBT right how to make a halwa will start with chat GPT right and and because of that it's becoming quickly your gateway to the internet your gateway to what it what you want to do the difference though is that it's not a dumb gateway it's a smart gateway in the sense that it understands your intent it can modify responses according to your personal tastes your personal beliefs and and it will give you responses which very very customized to you to some extent of course search also used to do that but an agent is smarter because essentially it's AI and it can do go and cover a lot more depth that search never could so search could always show you the things and then you have to do it yourself an agent a chat agent is smart enough to do most of it stuff on its own as long as you give it the right tools and that's where the world is headed so >> if I were to answer your question couple of years later and we don't know how quickly this will happen because things are moving really fast I could open a chat window and say, "Hey, I want to make halwa today." It tells you my recipe. It tells you the way I would like it because it knows my past habits and past behaviors. It will tell me that you like a bit sweet, so maybe you should add this much more sugar and this much less sugar and so on. You are very focused on your diet right now, so maybe you shouldn't add this much oil or ghee. You should put it you this is a way you could make it healthy. And then you say, "Okay, what's next?" And it tells you, "Okay, these are the 10 things you need. I know you typically order from Swiggy, so I've already created your cart on Swiggy. I know you prefer this brand of ghee and this brand of of floor and whatever. And then it adds all of it and says, "Hey, I know you typically pay this instrument. Can I go ahead and pay it for you?" You say yes and it pays it for you. And while I'm speaking in this usual flow, the complicated aspect of this is that it's extensible, which means I can do this conversion in Hindi. I can do this in Canada, I can do this in voice, I could do it in a third party platform that connects to Chad GPT. It could be anywhere. It's extreme extremely extensible and that what that's what makes it very beautiful. So I'm talking about it like I open ch but maybe 2 years out I'm not even open anything.

7:50 I'm walking through my home and I'm talking to my refri smart smart refrigerator or my smart like a Google home device and say hey I want to make halva today and a Google home does all of these things. So that's the most amazing part that it could be extensible and be invoked from anywhere. The way I see this agentic commerce is kind of already unfolding, right? Like in China, I see how Alipe is doing it. I see how you know, other platforms are trying to do it and there have been campaigns that have been pretty successful. But, you know, just jumping ahead a little bit, when there are giveaways, people really get into agentic commerce. you know, they they make their orders, they talk to their chat bots or sorry, their agents which go out and do it, but then after that, the habit doesn't really stick. So, I'm kind of curious like how you think about the process of even getting to what you just described. It seems like there's a journey to there.

8:53 No, >> I think and that's a very interesting point you brought up, right? Because there have been experiments on this even in US and it hasn't taken off, right? So and it's a very valid point to discuss why do I believe it will take off in India when it hasn't taken off in US it hasn't taken off to that point to that extent even in China and the reason is it's I'll go back to the journey of mobile commerce right like mobile payments got launched in multiple markets together it picked up in India didn't pick up anywhere else and the reason is very simple India didn't have entranced habits of doing payments so in markets like US people already knew how to use a credit card people had entrenched habits that when I have to make a payment I take out my credit card and Right? Changing that habit was extremely hard. It hasn't happened even while India has been doing it for now for almost 10 years. It hasn't happened in US yet because that habit is well said.

9:39 So everything else has built a layer on top of that habit. Apple pays a layer on top of credit cards. But you couldn't really change the core habit that a if I have to make a payment I take out my credit card. same goes to agentic commerce. Now in spite of all the efforts that have happened in last 10 years and building commerce in India which is apps like Swiggies, Somato, Flipkart whatever and we power a bunch of them the people who have built habit of shopping on commerce in e-commerce in India still very very tiny right so India has almost 450 million UPI users India has regular e-commerce users less than 50 million regular e-commerce users and this is across all the apps right so there might be 100 200 doing one at once in a while but it's 50 million regular e-commerce users. So the e-commerce penetration is that tiny in a country of billion and a half almost. We are less than 50 million who do regular e-commerce shopping. It's not that these people don't shop. There's a mass massive amount of purchases happening on unorganized retail. Massive amount of purchases happening on offline stores and so on. It's just that people haven't yet built a habit of shopping on e-commerce. Right? Compared to us where that penetration is 70 80% you're still less than 10% in penetration. We look at data in Razer Pay. 70% of all GMV on Razer Pay out of the 180 billion that has said >> that's the gross merchandise value.

10:51 >> Yeah. 70% of that comes from just 15 million users five right that is it is that thin right and that is and so so while we might look it as a negative I see it as a positive because of because we don't have entrenched habits we are likely to pick up a new faster way of doing things better then somebody will say hey it's a 20% delta versus for us it's a it's a 50 60% delta compared to what we do now why I believe that habit hasn't changed in the overall world and why agent eiccommerce brings Does that change? The reason that habit didn't change is that the way Indians shop, we haven't been able to replicate that in online world. The way majority of India shops is we don't shop in supermarkets.

11:32 We don't like this plethora of choices where you pick and choose and you do your own research and you come prepared in a in a large supermarket and hey I want to buy this today. We are conversational. We are multilingual and we are voice first. Right? And that's why for example if I take example of insurance purchases 99% of insurance in India is purchased through offline brokers in spite of massive online brokers who are much less biased who will show all your information. The offline brokers is likely to be biased likely trying to push something that gives them more incentive. People still buy insurance offline because I don't want to compare these three insurance and see does it cover everything. It might not cover something that I want.

12:08 The display might not be in the language that I understand. I can't chat with somebody and say does this cover that one particular thing I want it to be covered right so so the voice conversational and multilingual is not solved and that is how that is what India needs to to do commerce online this applies to every purchase it applies to insurance it applies to wealth purchases it applies to commerce purchases it applies to almost travel purchases almost 50% of travel purchases in India happens through small travel operators because people still don't trust that hey yeah a make my trip or a clear trip shows everything but I don't know if I I I don't trust myself that I am buying the right thing so who do I ask right and that is what agent e-commerce solves for so so a that transition never happens so there are no interest habits we are fighting and b the the natural way that Indians want to shop is actually possible with digital commerce which with the kind with the digital commerce we have built so far hasn't really replicated it and I think because of those two things I feel it's a bigger opportunity for India than most of the world and the experiments that we ing in most of the world today.

13:12 >> Interesting. I will also say hersel that this 15 million that you said is something that I mean I had written about this long time ago where I said that this is a common theme across when you talk to like e-commerce players. They always say that it's there's a strong power law with respect to our purchases and I had actually put a term for it. I called it India's California users. I said that there's such a very very small set of people doing this. We started this by just jumping into all of this futuristic stuff and I think I should also pull back a little and talk about Razer Pay. It's also a testament for Razer Pay that now you are a company where I don't have to really introduce and say what you do, how big you are because I think most of our listeners already know. But for the for the record, let me just say this again.

13:48 Razer Pay powers you know your FI25 revenue was roughly like 3,783 crores. That's what is public. I'm sure that there are numbers now that are still going on. Broadly 55 to 60% of India's online payment gateway market is powered by Razer Pay. In other words, you are essentially a huge infrastructure player at the bottom of all of this. I'm sure some of these numbers have changed, but these are public numbers for now. I will also say that and I think we'll jump into this now. We've spoken about this broad future. Now, I think we'll get into the steps and the sort of the milestones that you've been building over the last few months towards this and you can tell me how all of this connects together because I can see a pattern, but it'll be interesting if you can tell and I'm going to read them out first. I think it began I think the first major step that I could find was October 2025 which was your first pilot where Razer Pay you tied up with the National Payments Corporation of India which of course runs UPI and OpenAI and also you were basically able to create one part of this which is I want to order groceries from Big Basket in chat and so all three of these work together via UPI and you were able to do that and there was a demo that I saw about it and everyone knows that was the first major thing that happened. Then in Feb 2026 you powered agent payments on claude.

15:02 Same thing or similar thing but it was on Zumato, Swiggy, Zeppto. You don't have to leave the cont and you built it on UPI reserve pay which means that you don't have to put the pin. You basically select a merchant and say okay I'm going to put a certain amount for this. So you can just do that fairly seamlessly. It's a so it has consent right from the beginning. Then in Feb 2026, you built the MCP server and you partnered again with Zumato and also with Grani superu. And basically what it does is it generates a live UPI link or a reserve pay mandate mid call. I believe this is more of the Bhat thesis. So it's not just the 15 million you're going a little bit beyond that. then of course agent studio which built which is built on anthropics claude agent. So your agents are I'm just going to read out some of the agents I found.

15:46 abandoned cart conversion, dispute responder, subscription recovery. So you now you've started connecting to Shopify, WhatsApp, ship rocket, etc. Then there is the merchant side of it and now what you did was you basically had natural language where somebody as a merchant side can apply and upload a bank statement screenshot. They get instant reconciliation and that's integrated across claude code etc. And you also have inapp agent e-commerce pilots running with you know Zomato, Swiggy. So this is another surface where you're basically doing this. I'm just looking at this canvas. So it's essentially one is of course the chats and the you know the AI models etc. Then there is the merchants. Then there is the layer where you're helping merchants do reconciliation. Then there is within the apps itself. Then there is voice.

16:37 What are these building blocks for? >> Talk to us. How have you built this? There is something going on. a lot of action over the last six seven months. >> Yeah. And I think that's where the ecosystem has been in the last 6 months. We are trying to keep up with our vision of where that ecosystem is heading. I think the the broad idea is that that this like we believe that this there two years out we'll live in a world where most of your day-to-day operations are powered by agents whether it's inside the companies or whether it's for individuals themselves. And we're trying to build the layers needed to to support that world from from a commerce perspective. Right? So one is on the on how our merchants service their customers.

17:19 >> Okay. >> And and there and we are see we are an infrastructure company. So we not taking a bet where that world heads. It is possible that most of the commerce is initiated directly on apps like chat, GPT and cloud. So we are building connectors on that side. The the other side is that people still come to these merchants because they still entrenched habits. People still come to a Swiggy and want to order, but instead of searching and finding and all of that, they can talk to an agent inside Swiggy, a Swiggy agent essentially and talk to that that hey, I want halwa today. I I I want to make halva today and Instammart store can pick up stuff for you. And we are building both sides of that that is on the merchant side. So wherever a merchant a business wants to service it customers whether it is through charge GB whether it is through any agent provider tomorrow or whether it's through an agent inside their own platform we want to be able to support commerce on that and the reserve pay protocol allows us to do that in a far more seamless way because the real value of agentic commerce will come in when an agent can buy stuff with just a yes or a no so that you can just talk to it and you're not opening your phone and typing a pin during a transition. So, I reserve pay allows us to build a protocol where people can authorize once and then then the agent can keep buying stuff for you without you interacting and exchanging payment information or opening your mobile phone every time.

18:35 Like I gave an example, people could be doing it on voice on a Google own device, right? like Swiggy has an MCP today that you can connect to any agent and then you can talk to it on voice and then it can do it for you. So that's one side. The second side is how do I make my my merchants and businesses themselves more AI enabled in the sense that when all of these payments and the volume of these payments volume of these payments increase they're not deploying more and more people to manage these payments and that is 18 studio where I provide a way for businesses to build agents on top of Razor Pay to manage reconciliation manage refunds manage recovery a customer drops off midjourney my agent an agent can spin up offer the customer a discount or whatever to get him back all of those things. So those are the two two broad sides of it and they are of different favors in each of them.

19:23 >> So I think the interesting part for me is I'm just talking from a product standard because I mean the product is what is most fascinating for me. It seems like all of these at some level are complimentary but also contradictory in the sense that at one level you are trying to basically say look habits may change and people may go let's say hypothetically they may abandon a swiggy or a zomato etc and start making these purchases on say a charge or a claude we're going to help that or maybe they're going to continue doing this on a big basket or a swiggy or a zomato or a zeppto wherever it is we're going to help that maybe these merchants are going to essentially need some help in order to do reconciliation. Sure, we are going to help that. It feels like in a sense you're hedging on all sides because you don't know which of these futures will play out.

20:10 >> Yeah. And nobody knows >> and that's deliberate. >> And that's deliberate. And I think we want to bet on the future that the future is going to be driven through agents. But what shape and form it takes, I think it's very early to call that right now. >> As an infrastructure provider, like for example, when we build payments, even pure UPI, we don't really care. We didn't really care whether the payments happens on web, mobile, whether it happens through links. We built all three and honestly today all three happen. It h a lot of payments happen web mobile links pages multiple products that we built and people different markets and different segments of market picked up what worked for them. And similarly in the agentic world different segments I don't think it's going to be either or it's going to be all and different segments of the market different flows will be better for them.

20:52 For example, when you're doing deep travel planning, a broad-based approach, a chat GPD will be a better interface. But when you have a specific thing to buy, a Swiggy will be a better interface or a or a blanket will be a better interface. Hey, I want to buy like this month's ration. I don't really I'm not really researching on it. I just want to buy these 10 things. So, I I will open an install and just buy it from there.

21:12 >> So, the analogy is like it's almost like there is a human body and there is a nervous system underneath. And what you're doing is you're saying we're going to build the nervous system. We don't know what shape the body is going to be. We don't know what parts of the body is going to be used for what. We're just going to make sure that there is sensory this thing all across the board. Is that like a fair analogy? Imperfect.

21:30 I'm not a biology student, but broadly that's the analogy. >> Neither am I. But that's a that's a fairly good analogy. >> Okay. Fair. well actually I've been thinking about this from the standpoint of say small and medium businesses, restaurants, mom and pop stores that may be online selling the products that they stock as well. And I I can't help but think that, you know, when agentic commerce happens, they kind of lose the relationship that they often maintain with their customers, which is actually really essential for SMBs. And you know this is leading me to the conclusion that agentic commerce really favors existing platforms. And there's this scenario where a lot of small businesses could get wiped out really easily. You know in the same way that Walmart wiped out a lot of mom and pop stores >> and that's a very good example. I I have a very kind of exactly opposite view of that. in the pure internet world when internet is winning against offline which is happening in general. I think there that part is true whenever wherever e-commerce is penetrating deeply and Amazon is becoming larger or Walmart is becoming larger and the small and medium shops are losing out. So that transition is already underway has been underway for last 10 years that things as things are moving online they're moving towards organized. So when in offline world I would go and order in a restaurant but in online world I order on a Swiggy or a Zomato because in online world discovery is hard. So I have to go to one or two platforms because the cost of discovery the CAC is extremely expensive. No restaurant can continue to spend on it, right? And that's that's the world we have built over the last 10 years because you can't there's a cost to that discovery that you have to keep spending on. In an agentic world I feel it's going to reverse its course a bit because an agent doesn't really care about entrenched habits. It doesn't really care about disc. The reason you have to pay a lot in the older world is because a human has entrenched habits. A human is not going to open up every day and say now I want to eat a eat some different I want to eat a halva today so I'm not going to find a new restaurant that is expert in ha >> you really like the halva example >> because you started >> no but that's also because I think you use that example first on stage somewhere so you put it in my head so please continue on >> yes that was a good example so the human doesn't it's very hard for humans to every time pick and choose yeah I bought on Amazon three times in the last week so this this week I want to buy something else. I don't even compare prices. I just open Amazon and buy because it's too much effort to compare prices in three different places. Maybe the brand sells directly and cheaper than Amazon, but I'm not going to open the brand website. So maybe I eat, let's say I take buy Whole Truth, maybe take an opposite example, a whole truth protein, which is kind of an opposite example to halo, right? And I buy I buy from Amazon. So I keep buying from Amazon. Even though whole truth on its own website is selling much cheaper and with much better offers and all of that, but I as a human, I'm just habitual to this. it's too much effort for me to do it for every purchase. An agent doesn't really care about it. An agent can can continue to make that effort. It doesn't shy away from effort that a human does.

24:34 So every time you're purchasing, the agent will try to compare what are the best options to buy it. What are the best ways to buy it? It will still latch on to performances. So it will still latch on to when people say, "Hey, don't buy from there's a new website. Not enough people have brought from it. So it won't maybe buy from it. But if enough people have brought from it and say hey it's better to buy from who directly because they give something better and don't buy it from Amazon the agent will actually do that change change the decision and recommend you that hey buy this from whole truth because that's what people have said online. So there's a possibility that the entrance behavior can get unlocked and broken far more easily than the online world. In the internet world it was actually hard and most of the companies were built on buying that entrance behavior. That's why people spend a lot of money on CASS and all of that that once I buy this entrance behavior, it's going to be very hard to change. So I could start an Amazon today and I could be better than Amazon in quality and servicing. I'll still spend billions of dollars to change that habit that people have of Amazon versus I could start an Amazon today just for Bangalore. I could be better than Amazon and Bangalore. And once enough people say that I am better than Amazon and Bangalore, the agents in general will start discovering that and start buying from me. So make it easier for newer businesses or smaller businesses to fight as long as they can deliver better performance or better outcome than the larger guys. In the older world, it didn't matter even if you had a better performance because nobody's really checking your performance every time.

25:51 >> This is interesting because it's a similar argument that Tik Tok made when it was really making a push into markets like Indonesia and it onboarded a lot of SMBs. Whether that worked out is still unresolved. >> No, I understand. But I feel like this is I mean I get Hersel's point. It's a little different but okay now I'm going to push hersel on one important question which is that I feel like you've you've put this very interesting future and I buy this argument broadly speaking so I think I'm going to ask you now in this future who really loses you cannot say everybody wins because that's not true right it can't be that everybody like okay let me put a couple of hypothesis on the table you should don't have to agree with okay I think two sets of people broadly seem vulnerable let me just say vulnerable and not necessarily losers because we don't know I think number one vulnerability comes down to anybody who has created a entrenched position as a result of being sort of like the default option like maybe it's the Amazon of the world or the zomats or the swiggies or the zepptos or whoever it is which is like anything that is like a default behavior where I just go there blindly I feel like they seem to be a little vulnerable that's one I think the second kind of vulnerability is anyone who has gotten a certain kind of an advantage through expecting and believing that customers and users will basically it's too much of a friction to do the other thing.

27:10 Right? So, as an example, hypothetically, there was a company, I'm not going to name a specific company, but let's say there was a company that was essentially thriving by adding this hidden charge or a convenience fee or something like that, right? Which I'm sure that you probably have much better visibility than me, but who does that? So, I feel like those kind of companies and merchants because now I can just tell an agent, no, find the cheapest thing, make sure that you find it. So, all these dark patterns, etc. Sort of like don't get seen to agents. These seem to be two kinds of companies but I'm sure there are other and deeper ways how you've thought about it.

27:41 >> Yeah, I think the macro view would be anyone who's anyone who doesn't keep up with the changes that are going to happen is definitely going to lose. So anyone who's like in the past 10 years a lot of companies had built massive like put a lot of effort into first acquire these customers build these habits and all that and then over time said I don't want to innovate anymore because I have this pool of customers who are going to come to me no matter what I can just keep earning more from them >> I think that that is going to get accepted at the end of the day it's the customer who wins and the customer is going to be able see what happened with internet is democratization that a customer who could only buy in his neighborhood store can now buy every from anywhere in the in the country or maybe the world, right? And that if who won in that race? Did the neighborhood guy lose completely? Did the internet guy won completely? No. A lot of people changed their ways around. The customer won because now as a customer, I got cheaper products because I can compare the global marketplace. I can compare the entire national marketplace and then choose what I want to buy. Similarly, when agents come in, I can now have an intelligent agent that compares between those marketplaces as well and find me the best outcome there. So companies who keep up with this change and adopt it can really drive a lot of value out of it. I'll say so in US this battle is being fought a lot like companies like Walmart and Target are opposing it because they don't see that agentic commerce is going to create any new value for them. The market is GPE with e-commerce 70 80% people already use e-commerce. So there's no new customer coming online because of agent tech. So it's a fight between channels. But India because of the original argument I made because we still at 15 million doing 70%, less than 50 million doing regular shopping there's a new market being created if that number moves from 50 million to let's say 100 million which is Chad GPT DAO today right it's a massive unlock for everyone across the ecosystem right because now no more people are going to buy online more people are suddenly on e-commerce there's a massive unlock that's going to happen the offline guys make their products more available online more available on these digital platforms can still fight for that same race versus fighting in the local. So there is a massive value to be unlocked. At the end of the day, it's a customer who will win. So somebody who services better, who performs better, who gives it the cheapest cost and is able to keep up with these times is going to win. Anyone who doesn't meet all of these criteria and is just living on the on what they built their outcome out of in the last 10 years and say, "Hey, I'm going to keep earning more. I'm going to keep squeezing this more. I'm going to lose."

30:01 >> I'm still going to push you, Hersel. You basically said customers will win, which I which Okay, fair. Nobody can disagree with that, right? Let me just push this a little bit more. I think for instance, let me just say few other possible winners, right? I think number one, I think one set of people who will definitely win are logistics players. And this is one of my thesis which is that I feel like anybody who's doing logistics right now because logistics as you know of course is like a game of just inches and rupes. Every rupee, every single margin matters. So what's going to happen is right now those margins are probably accumulate and acrew more to the merchant side especially if you're big income. So hypothetically if you are someone hypothetically like a flip card a large percentage of the value occurs to you and something like a delivery or a ship rocket or anybody else gets like a smaller set of it. I think now with agents what happens is there is an opportunity to bypass these altogether.

30:50 So that's one. So I feel like logistics companies there's a huge opportunity over there. Number one, I think that I I think that the people who are actually going to lose are going to be this set of marketplaces that don't offer anything beyond just a catalog and discovery. I think we all know like for as an example if you think of well okay maybe not a flip card but I think if you think of I'm sure there are many of these set of these market I can imagine like the shop clues of the world or any of these places where there is just a matching of oh here is a place where you can see everything. Let me let me put add there right like see I think a lot of that particular in the offline world will actually be very true right so there are lot of massive offline stores without naming them which the only thing that they do is bringing the products and services available in a coa fashion so you can do discovery inside the store correct right and those will lose because the transition to online will displace them first and the transition to online is going to accelerate because of agent tech so it's going to displace them first in online marketplaces is you're right longterm there might be a fight but in short term because most of the new market places in India are still fairly new it's still fairly innovative the the increase of market is going to impact them more than the disruption on channels the channel disruption is very tiny if that increase in market happens from because you're sitting it's so tiny like 15 million doing 70% yeah maybe 2% of it goes away maybe 20% of it goes away it's still very tiny the uplift from that 50 million regular shoppers going to 100 million is so massive that that 5 and 20% channel conflict is not meaningful and that's why our Indian brands are not fighting it the way US e-commerce companies are. They are not worried about channel if the market size goes up. Who's going to lose is in the offline world people who make a lot of money because they're sitting in the right neighborhood. The this neighborhood doesn't know how to buy online or gets very confused buying online. They want to go to a store, talk to a guy and buy stuff because that's the only way they can answer their questions. that is going to get disrupted very easily.

32:53 >> Interesting. I mean this is a little counterintuitive but by the way I will also say for our listeners that the one part where I agree with you on the online players not fighting this is that like of course we know that Razer Pay and a couple of other companies are also like they really like put their you know they've really st their positioning on agentic commerce but one of the companies that's also done that is Flipkart like Flipkart has been very aggressive in stating that we are going to be ahead on agent >> all the new age players Zomato is putting a lot of effort Misho is putting a lot of effort as I said because these are like our our new age players are fairly new age and they are all trying to innovate a lot. So you can't compare them to a Walmart or a Target in the US.

33:32 >> Okay. >> Okay. this has been a very leftrained conversation as is often the case on zero shot. I follow the logic you know even if I'm skeptical the logic seems mostly sound right. but I'm also trying to think in the the shoes of somebody who's not within those 50 million people who are on board with e-commerce. And going back to the voucher example I mentioned in China, you know, that works not just as a marketing ploy, but because it's a language that anybody can understand, right? So Harshel if you if you had to I guess communicate this to somebody who isn't quite you know forward in the you know internet economy you know what is the threeline five line sales pitch there >> yeah the sales pitch especially for the Indian customer and I'm talking about that particular Indian customer who has not transitioned to e-commerce today and and it has not he has not transitioned to e-commerce today because he prefers to speak to somebody before he buys something. That's my base reason for it because most people have smartphones.

34:45 Most people are using UPI. They're still not using it for commerce because they want to chat with somebody. And I believe agentic commerce allows you and gives you the exact same experience that you're used to in the offline world on a digital world. Like to give a specific example, IGO launched this pay travel assistant. It essentially is that that you go to a today's shop, you stand in a line and you talk to a travel broker or travel agent who the only job he's doing is you say hey I want to buy a flight from Bangalore to Bombay he's opening up an online platform searching it and telling you hey this is the price and you say is this all of this covered yes all of this is covered will will it meets all my requirements yet it meets all requirements and books it for you right why can't you do that now in an online world and that's what exigos travel agent does for you or I can give an example of Misho right Misho this AI assistant now where you open it up and you say hey I want to buy this and I want to buy this blue color t-shirt that's within 200 rupees you can find and exactly tell you that this is this is how you shop in a store this is how most Indians shop in a store today that you go to a small store say I want this color t-shirt this is my budget and tells you exactly this and then you can ask him two questions is this cotton is this right fit does it fit me right and all of that and you can buy it and a AI agent allows you to deliver that exactly in that Okay, in exactly the language that you want. I think that's the core pitch to somebody who's not yet jumped that bandwagon of doing on e-commerce. And I believe the reason they not do they don't do that is because they need a certain experience which is not possible with agent e-commerce. Okay, last question from my side which is that suppose I asked you from razor pace vantage point. What is the one thing you're waiting to that you think is going to happen next or you expect to happen next? I'm not going to ask you when it's going to happen because nobody knows. Maybe it's happening in 2 months or 6 months or 9 months, but you feel like the next meaningful milestone and I'm not saying necessarily that Razer Pay has to crack it. I'm just saying that the minute you say once I see this happen, that's when you know the game has gone to the next level.

36:45 >> I think personal agents is one big change. which I think no company has cracked it today. like open claw and Hermes were the tech examples of it that somebody's very tech aware can do that and that's so massive way that happen wave that happened in US but still not generally available not everyone can just spin those up some company's going to figure out a way to make personal agent available for you which is like every person can have an agent that almost operates like an assistant for you and does every day-to-day task for you find me the find me a a a flight from Japur to Mumbai at the cheapest cost make itenary for me and all of that. Yes, you we are doing it on chat GP today but then it's broken. I do something on chat GPT then I jump somewhere else and do rest of it.

37:28 Somebody maybe it's open AI maybe it's entropic maybe it was somebody like Manis somebody will solve it that a personal agent for every individual which understands knows you builds memory and every conversation it becomes your go-to agent for everyone it almost becomes like your personal assistant in that way somebody will crack it and that's going to change the world quite linearly feels like I mean we have none of us have said this but feels like this was exactly what Siri was supposed to do at some point in time I mean you're talking like a Jarvis kind of a thing by the way I I said it was my last question But I have to ask you this one thing because I know that you are really big on tinkering and you use this yourself. This is not a theoretical thing. I have been seeing following you online. can you tell us a little bit about what kind of agents you use and what are your use cases are these days?

38:12 I know you have something like where you use it to like summarize some podcasts and articles. I know that. Apart from that, what else are you using it for? What's the latest coolest thing you've been doing? >> A lot of stuff. So I have a open claw and armies both agents running in parallel. They kind of act like my personal agents. And can I can I assume that you built it yourself for the most part? >> I mean the agents the platform is available as it own but I built like I customized it for my use cases myself.

38:36 >> Correct. >> Right. So I tinkered with it every day and it does a lot of like both of them together do a lot of stuff for me. I have two different agents. One is a personal one is a work agent and I wanted a separation. So so the work agent like does before every meeting it will summarize the conversation I've had about this meeting. Who am I meeting? Have I met them in the past recently?

38:53 What was the last conversation with them? any minutes of meetings in the last meeting that would need to be carried forward and give all of that. So before every meeting in the morning it tells me what my calendar looks like any conflicts that are there if I should I can send it a message and it can resolve those conflicts itself and so on. So that's my work agent more interesting stuff that it does is we have this concept in raise where we have staff meetings where every month like the different teams present what is happening and what next to the sea. Now one interesting thing and somebody might not like it is teams tell you that hey this is going to happen in next three months next six months and so on. When it happens everybody comes and tells you this happened. When it doesn't happen people typically avoid talking about now my agent keeps a track of promises that that were made and has what was the status is the status of this promise updated or not and then I can because I won't remember but the agent will remember it for me. So that's something very interesting and unique that it solves for me. anything on the personal side that you want to talk about that you have built an agent for?

39:50 >> On personal side, I think that my agent now correlates my health from my whoop, my CGM, my sleep and my food habits. It correlates all of them and gives me next day how the yesterday went and what worked, what didn't work. For example, like some general insights everybody knows when I sleep late, my recovery is poor and all that. But it can correlate stuff. So for example, I order a lot from Swiggy typically. So it can correlate and tell and it also pings me every 4 hours asking what I ate. So now it correlates and tell me that when I eat Indian in the night my my recovery is generally very bad and my CGM response is fairly poor. My sleep quality is worse. When I eat typically when I eat eggs or Italian in the evening my sleep quality is better and which was weird that Thai food for some reason doesn't suit well with me in the night. So I need to be careful about it.

40:35 So that has been very interesting that it specifically tells me what kind of food habits and what timing and all of those work well for me. And as more and more data gets fed into it, the better it gets at predicting this one. >> So what you're really saying, Hershel, is that the next milestone, you said that personal agents, but I actually disagree. I think the next milestone is when your agent tells you that halva is bad for you.

40:56 >> I don't think our agent used to tell me that. But yeah, >> thank you so much, Ash. Such a wonderful conversation. look forward to whatever is going ahead and I have a feeling that whatever happens with agility commerce in India I think that razor pay is going to be involved in it and be the infrastructure of it that was zero shot the Ken's weekly podcast on the biggest developments in artificial intelligence our hosts and commentators are Pravin Gopel Krishnan Rohin Dharmakumar and me Brady our sound engineer is Rajie Cenne who makes everything sound spectacular don't miss our zero columns which are published every Saturday. We'll be back with more for this podcast next week.