edited-audio-rahul-chaudhari
===

[00:00:00] All right, Raul, how you doing man? Thanks for coming on. Doing great. Thanks for having me. Excited to be here. Yeah. I'm excited to have you. Between Kohl's, Amazon. You've got almost the last decade covered pretty strongly in retail and then Airtel, HSBC. Before that you've been all over the place. I do wanna talk to you kind of about how you think about product and how this like depth of different kind of industries have played in first.

Before that, let's just like go all the way back. Do you wanna give us just the 92nd TLDR on? How did you end up here and how'd you get into product and that kind of piece. Absolutely. I actually did not start in product, I started in marketing. Yeah. You know, big food. I was really excited, really drawn to consumers and brand building and marketing and that's where I started.

I kept my teeth in that worked in a CPG company building brands. I used to handle a food, a biscuits brand back in India and a couple called Britannia, which was part [00:01:00] of, or sort of associated with Danon. Grew back then. And then my curiosity then took me to more p and l roles because I wanted to also grow as a business leader.

You know, connect the doors to marketing and brand building to the business impact. And I did some p and l roles in telecom in a company called Airtel, which was the largest telecom company in India at that point of time. Fourth largest in terms of the subscribers. So it's a fairly big portfolio in that sense.

And those were sort of p and l growth roles before, you know, formally they were called growth roles in that sense. Yeah, and then I joined Amazon in 2015 and that journey took me to building products and the regime technology to sort of in a scale part. I had done product in a CPG and a telecom world before, but you know, product in the true sense of software and digital is really where I take my teeth at Amazon.

So in that sense, you know, sort of grown and how my impact is at the of marketing of business p and l and product. In that sense, if you look at where product is going right now. You kinda have to have the knowledge across all three. You maybe not as [00:02:00] deeply as you went where you had senior roles and all these things, right?

But you at least have to understand 'em or else. And I think we're very quickly with AI and how fast things move and just the furthering of agency gonna run into the half and have nots there. Yes. Where some people understand these things and understand go to market and how to distribute and how to think about the ledger in profit loss and the financial aspect, or you're just like a feature kind of person that's probably be exponentially less important.

Absolutely. And the real at is, you know, you've gotta understand for any problem that you're dealing with, right? And the way I started is this desirable from a customer perspective, then that's where I really start and begin every problem or every idea that I have. Then I look at, is this really viable from a business perspective?

Does it make sense? And then you look at the technical feasibility of it, and that's where my product and technology understanding comes into play. And the question that I started asking myself now is, why now after all of this. And why now could have different nuances to it. One nuance could be this makes sense to prioritize right now in my roadmap, and therefore why now?

But it would also have some tailwinds, which make it much more [00:03:00] sort of ideal to do this now versus having done it earlier or doing it later. And that's really the lens I've tried to apply across my convers. One thing you said that I had seen written and I thought was really interesting was the concept, and I think you'd be hard pressed to think of this without spanning at least marketing and product.

Mm-hmm. Is this concept of like, regardless of if it's your digital banking experience or Amazon or anything like that, they're the same customer. So why do we expect people to put up with a mid experience on your banking app and expect perfection from Amazon? It's the same person. They know what great is.

Exactly. And I'll give you an example, right. It was a few weeks back, I was trying to bundle my insurance. So I had who auto insurance with company A. I had my home insurance with company B. I had my landlord insurance with company C and it was becoming crazy to sort of keep track of everything. Obviously I bought those insurances at different points of time, but it made sense to bundle them, maybe even call from them so that, you know, tracking that becomes easier.

It's one place, it's one app on my phone versus having to do this in different [00:04:00] places, four different apps, that kind of stuff. Right. But that journey itself was a little bit of a pain from a consumer angle at least, right? So for example, here is what is broken for me. One was the intent was very clear on my end.

I very specifically said. This is what I want to do. These are my policies. I want to change some terms. I want to co-term them, but there was a very clear coordination, miss, right? This was not an underwriting edge case at all. One example, right? So if you go to different marketplaces and want to buy insurance, abundant insurance, I found they only allow you to bundle one or and one home.

But what if I want to bundle more than two, at least had three insurance requirements. I wanted to add an umbrella cover on top of that, but the initial bundling requirements wouldn't allow me to do that. That didn't spun into a conversion where I can now speak with every company. You know, some companies require me to go through an agent.

Even when I started my discovery online, the agent conversion was back and forth or emails or slack or messages. It took me almost a month to get this stuff. Right. [00:05:00] And you're not a digitally unsavvy person. Exactly. Yeah. And my sort of experience there, both from a consumer but also as a product thinker here, is why should I, as a consumer have different expectations from my digital banking or my insurance company or my app on the, you know, fitness based I go to as compared to when I shop, say e-commerce.

Right. Ultimately, when I start shopping or start sort of discovering a certain level of experience. I start building those habits. Mm-hmm. Now, if you're gonna force a customer to build different habits and different experience reeducate them, that adds a tremendous amount of friction search. For example, right now, if you're used to doing search online, you prepared know what to look for, how to look for, et cetera.

Imagine now you go to a different site where search does not behave the way you expect it to behave. Now that's a relearning process for you, and the point is, why would you want the customer to go through that? Because the same customer another day, right? Customer has similar expectations and their bar is raised similarly from one level to the other.

So if I [00:06:00] have a great store experience at a Whole Foods or at a Trade Joe's or at Kohl's, I did not expect a similar great experience at any other store environment that I go to. Why should that be any different? I'm curious to get your take on this because the positive ways, if you have a great experience, you will expect that everywhere and as you should, but separately, I've seen a lot, and this is probably more common on the B2B side, even when you have a bad workflow or a bad experience and there's a better option there, but it requires a change in how you do something.

People get very used to that kind of, this is how I do this, and this is something we've run into here is like, so I don't know how familiar you are with what we do at Log Rocket, but basically we start out doing session replay. We do basically AI agents that watch it for you. We have agents that kinda read all the feedback and the idea is you don't ever have to watch 'em.

It just tells you what's important. Sorry, and I talk to users a lot. Run into people going, I used to search and to find things that I've heard about from other people and go look at what happened. Like, no, you don't need to do that. Like, that's the entire point. But because that's the workflow [00:07:00] and they bought it for the, for the proactive, but it's just the workflow that people get.

And so it's interesting to see like you get locked in and when it's a great workflow, you expect that everywhere. When it's a mid workflow, you still kinda expect that everywhere, and breaking that habit of people is really, really hard. Now imagine if it's really positive, now you have double reinforced.

Absolutely. So whether it's B, C or B2B, that aspect of project growth, which is one I'm trying to do more friction as much as possible. Second is I do not want to reeducate you, so you have to reinvent that wheel or sort of relearn something. I want to leverage what you intuitively know so that it becomes easier for you to use my product.

And maybe this is a good story because one of the major places you've been, you spent almost a decade at Amazon, right? Seven and a half years at Amazon. Yeah. So you did a lot there. But one thing I found really interesting is you actually had a part in what is a major workflow for hundreds of millions of people.

The the actual homepage, right? Right. Yes. So let's dig into that. I think there's an interesting story there of like how do you make these changes and add positively to this kind of like [00:08:00] workflow and how do you help people have a great experience but also improve the bottom line for the company and everything.

So you can add to both, you can have positive outcomes on both sides. Yeah, absolutely. So is very interesting. And what was my. More sort of, I would say, peak experience at Amazon, doing something at scale and then launching it and scaling it globally. The story is really interesting because this talks to both Amazon's philosophy of what Amazon wanted to do on the homepage.

Mm-hmm. Also how it wanted to be customer backwards at the same time. What does customer backwards mean? Yeah, customer backwards really is you start from the customer and then work backwards of that for anything that you wanna solve for, right? Instead of saying, Hey, I have an idea and I wanna do this, which is solution forward.

The problem there is if you already have an idea, then you already, you know, sort of tied to it, you are ready to it in some sense, and your ability to look for other ways of achieving the same outcome could be limited. Whereas when you start from a customer and work backwards of that, then you are in some sense solution agnostic.

You are trying to figure out the best way of [00:09:00] solving that customer's problem and therefore customer backwards. And I love that because I think it's a very Amazon thing, but clearly, I mean, maybe something more people should mimic is if the customer has a good experience and is able to get what they want and feels good about it, it's probably be good for the business long term.

So how did this work from that standpoint? Like what was the customer goal, I guess you guys were working on? So on the Amazon's homepage, you would find a rotating carousel, which Amazon used to call the hero. Most trauma, uh, experiences would have something very similar or a version of it right now.

Amazon's thinking then was we wanted to use the hero, which was the number one placement on Amazon, the third most visited website in the world. The number one placement there, almost like a waterfront property to drive what Amazon called the big bets. Mm-hmm. So Alexa, Kindle, or Amazon Prime, or music or pharmacy, or grocery, all of those programs had sort of an allocation on that placement.

Now the challenge was that each of those programs had their own metrics and they were trying to optimize for their own business in [00:10:00] some sense. So Alexia was optimizing for units, prime was optimizing for subscription. Music was optimizing for podcasts and streaming. And grocery was optimizing for getting customers adopted into the grocery behavior on Amazon, which is not the first place you go to when you wanna buy grocery.

Right. And the challenge really was that. So in some sense, in different businesses. Trying to apply different rules and different metrics and different optimizations, but in a way that was becoming not customer backwards because we were not starting from what the customer wanted to see in that placement.

We were starting from what we wanted the customers to see there, which is Amazon's big bets. The other challenge there also was some sort of an over exposure. Now this is fairly easy for you to understand. If you sort of look at, you know, watching a television advertisement, for example, right? It's the first time you watch it, you might, you know, see it, you might not see it, you might be interested maybe a little bit.

Second time there is some hook. You know, third time you get the message, probably. But if I show you the same message, the same ad, 7, 8, 9, [00:11:00] 10 times, what I'm doing is one, I'm wasting money. And at some point you've already started not paying attention to it, right? And if you keep doing that again and again, that really drives a bad experience with the customer.

So we are sort of seeing some of that on the homepage as well, where all exposure to these programs was causing what we called a zero blindness. So customers were ignoring parts of those. So what we really did was we did three things. We looked at the policy, which is eligibility and exposure gaps. What we then look for was measurement.

So we solved for common measurement across all of these programs. So it is stick to whether it is advertising or external devices, or we are looking at prime subscriptions. All of those programs have brought them to a, a common set of metrics or a common set of measurement across perception, building, habit driving, and economic impact.

And then what we did was we put some governance into it in terms of defining what really a campaign means because different teams have different definitions of a campaign. What does a creative mean? What does exposure mean? And then put guardrail so [00:12:00] that no one program can. Take, you know, a bunch of the exposure, but we do it in a way that is customer backwards.

Because once we start seeing signals from the customers, if they are engaging or not engaging with that program, then we show them something else, which is the next best action or the next best program that they could look at. So in some sense, we really started looking at it from a customer perspective, but to really tie that to the business impact as well.

The net effect was we reclaimed some 41% impressions that were wasted because of work exposure that led to, you know, hundreds of millions of dollars of incremental business impact for all of these programs. So how were you able to kind of come to an understanding of what the customer was looking for there, or how did that work to figure out what they wanted and what was working, what wasn't?

Was it people clicking on stuff? Was it signing up for Prime video or something? How'd you kinda discern those things? One, we started with a lot of experimentation. Actually. We ran about 20 plus experiments in five or six countries at all different programs to really understand what that placement was good at and what it was not good at.

Right. [00:13:00] So we sort of understood that displacement was great at driving perceptions, but maybe not so great at driving lower funnel metrics. So experimentation obviously gave a lot of those insights. We also look at, like I said, the exposure limits as well. So data told us, there was obviously a very clear point of diminishing returns from impressions showing the same contenting, the same program beyond a certain number was actually flattening the incremental impact curve and in some sense, actually driving the customers away from that program.

So it was actually harmful more than being useful. So we'll cut some of that data. And then module a few things again to see what customers are looking at in terms of your question on what metrics that we look at. We really created that TriFactor of metrics across, does it impact perceptions that we care about for that program?

Does it breed habits and drive long-term adoption for that program, not just a short-term click or a short term purchase? And does it drive an economic impact, which Amazon measures in terms of free cash flow, not [00:14:00] just revenue today. But what is the free cash flow of you doing something on Amazon one year from now, two years from now?

You know, continuing to use Prime Video as an example. Yes. Is it like looking at, you show a number of impressions and down the line they have a higher rate of subscribing to Prime video, or am I looking at that wrong there? No, you hear that very correctly, but the one step that we would look at incrementally was are they just.

Sort of streaming on Prime video, or are they streaming on Prime video to an extent that then defines the adoption into that behavior? So for example, if you go stream on Prime video today for one minute, after watching something on, say, Amazon, maybe Amazon has a new show. And that's exciting. And you go stream and you watch your trailer.

That's great. At least I got you to sort of do that action on a different screen, on a different device. But that is not going to get you adopted into Prime video as a program, right? So every team, every program defines what adoption means for them. So adoption for a prime video or a Netflix would mean that you stream at least.

Three [00:15:00] shows for 30 minutes each in a period of one month, right? That's where you start getting hooked onto that program or hooked onto that platform, and then you come back again and again. Adoption for a grocery program could mean you shop at least maybe six times in a period of three months. But if you just shop once or twice, yes, that's interesting, but that does not get you really hooked onto that program or that behavior.

So what we care about is not the one first action. First Action is great. But what we care about is, does that get you enough into being engaged into that program so that you're adopted so we can then stop pushing that content or that creative to you, and then get you onto some other program. I think that's so important into how do you define success and how do you really double down things that work?

'cause if you don't have that specificity, you're just gonna chase. You may go, oh, this person, you know, if you just have a point, action. This person, we want them to stream. Yes. Well, if they streamed a trailer, right, who gives a care, right. Yeah. I guess looking at that, you start to get into, at a certain point.

Showing them more, they're just not gonna do it. [00:16:00] So you just, you're just throwing good resources after bad. Was that more what you saw? Do you think that's where you started to see like more impressions get hurt? The intention is just they were never gonna do it? Or did it actually cause 'em to not want to be like video customers in that case?

So, so a couple of reasons actually. And, and the way I found this is, you know, making sure your content, whichever the surface is action aware from the customer. So there could be examples of that. The customer is already streaming time, video, right. But we continue to show the content, so we haven't really captured the action from the customer and really realizing that, hey, the customer's already hooked onto this.

Either I move the customer to a new program or I change the content to this customer in a way that is personalized to where the customer is in that particular journey. So one could be from there. The second could be saying, Hey, you already seen, you know, any number of canteens or creatives, or whatever you wanna call it from prime video, and you are just not engaging with it.

Therefore, you have only given me some signals that you may not be engaging with this program. So either I figure out another way of reaching [00:17:00] out to you, right? Maybe the content is not working, so I experiment with a different content or content type. Maybe the program that I was promoting or the series that I was promoting was not to your liking and therefore you did not engage.

And maybe I could then figure out if there's a better way or different way to reach out to you. Or third could be that, you know, you just are not gonna engage with that and therefore I should stop showing that to you. Right? So just really understand the customer signals and then adapting to it. So it really depends on the program or the business that they're looking at, but essentially being customer backwards.

Are customer aware, one and second, being action aware. So you are not really dragging the customer on the same journey. You are aware of where the customer is and helping the customer move forward in that journey. It's such an interesting point where people could take that lesson from is you need to be aware of like where those cutoffs are, where you're just chasing bad after good, but you need to understand deeply what are you trying to get the person to do?

What does success look like? How do you not get a false positive? They did the thing and then where do you cut off the effort and move on? I love it. And then, you know, in this case, it's what, half [00:18:00] billion dollars in impact, which $500 million in impact it seems like, right? By just getting this stuff right for Amazon.

Exactly, because I'm not interested as an Amazon or as a co, well as an enterprise in selling you one stuff, right? Yeah. I wanna make sure I'm trying to serve you best based on your interests and preferences, and if you are giving me signals in terms of what those preferences are or are not. Mm-hmm. Then you know, I am in that sense, product agnostic, right?

Yeah. They want to get you engaged into that platform. Not forced you to buy one product or one type of service. You spent a lot of time at Amazon. You know, a good portion of that was testing and driving forward and, and how do you improve the experience for the customers. Mm-hmm. But a good portion of that was also thinking about the future of retail.

Right. Right. And so we are obviously in the middle of, in arguably, a huge. Shift in how things are gonna work with the rise in spread of ai. How does AI start to affect retail on both ends of the spectrum, right? There's both the tech native retailers, but there's also the traditional ones that are moving more and more.

Like what does that future look [00:19:00] like? Do they converge? Do we start to see differences in how these companies move, or what is just the general future of retail look like? No conversion is completely without talking ai. So I'll share my perspective and my observations. I could be wrong, but this is really where I have.

Drawn some of my learnings from most AI conversations. Start with, you know, which tool should I buy? Or what is our AI roadmap? Or what is the coolest AI tool I encountered last week? Or are we in an AI hype or an AI bubble, et cetera. Those are all interesting questions, but for me, the better question really is how should our business work differently in an AI world?

For me, tools are replaceable. Business models, how you create and capture value is really what compounds. So I'm looking at this in a very. Agnostic way, if you can sort of term that. What I would actually encourage anybody who at this was to really look at the business model canvas and draw two columns.

You know, one without AI and one with ai. And in the AI version you can go box by box to really say [00:20:00] what really changes in this new AI world. Let's look at an example, right? So let's say value proposition in an AI world, you can personalize a lot. So the question I would ask therefore is what new promises become possible when experiences are far more personalized?

They are more predictive, they're more conversational. So for example, can I now promise the customer, you know, right, for them bundles. Can I offer assisted shopping in a digital world? Which was, is it not possible or very difficult to do without AI or AI capabilities? Can I explain some of my options a little better to the customer and promote to them, for example, not just a product, but say an outfit, right?

So those are really the questions I would ask, and I would really go point by point on the value prop on the customer segments where I can now sell profitability. The channels on how to search, email, instore, et cetera. Now change when I can capture that intent far more better in a natural language. My revenue streams, my cost structure, you know, [00:21:00] what gets sort of permanently cheaper within AI world.

Can I now bring down, for example, my contacts per order? Can I bring down my cost per acquisition, et cetera. And those are really the questions to me that are more intuitive to a business level, which is new revenue high LTV or low cost. Really stop working that isn't connected to any of these aspects of the business model exchanges.

It's an interesting perspective. 'cause I think kinda what you're getting at is the problem doesn't change. Like retailers, you're trying to sell goods to consumers and ideally the more you can sell to them, the better off you're going to be thinking about this, like to double click on it a little bit differently.

Mm-hmm. If you think about like a homepage, for instance. If I'm an e-commerce practitioner, which we have a lot who listen, right? Like concretely, what are things I can start to test around here that I couldn't do before? Right? And the beauty is with the customer journeys of shrinking. With AI agent shopping, my hope is today is any surface.

My hope is to be outside of my [00:22:00] ecosystem, outside of my site, maybe a chat GPT sort of surface becomes my homepage, and that's really where the customer discover me, my products. If I have the integration on checkout, would actually buy from chat GP itself and never come to my website, right? So you gotta also think differently about, you know, what your surfaces today are and not just rely on the existing surfaces in a pre AI world.

That brings a really good point. 'cause like one is right, you can personalize better on site if you have people who are logged in or you know, whatever, they can come and you can have a profile for them. But yeah, there's the other piece, which is one age agentic shopping, you know, modules that will come and shop for you.

Or two, given where a lot of the models are moving and a lot of the companies are moving right now, you can just check out on chat, GPT. Like if I'm a retailer, how do I need to think about changing that? Like I don't control the experience anymore. To me that would be terrifying. Is that right? Should I be scared as hell or is this a good thing?

No, I think you're gonna be excited about it because there's so much you can do more on a new surface in a very different way. The way I think [00:23:00] about this is, you know, let's an example of search, right? In a traditionally commerce search. The customer has to exactly know what they're searching for. So in some sense, you're asking the customer to type in an answer by saying, for example, I need a red party address.

And that's when you would sort of look at that intent and you know, show the customer options of red party address or maybe some other options that go along with it, et cetera. But when an agent did search, for example, it allows the customer to state their problem and their context, right? The customer, for example, say.

I'm going to a party. I already have these four varieties of grace. I do not want to repeat those patterns. I have, avoid them earlier. The party is gonna have X, Y, and Z from my office. Some of my friends, this is what I want to show up like, and therefore really capture the intent, which is the question, not the answer.

And let the agent search therefore, sort of, you know, help the customer really find what is most appropriate for that context. So for me, when the experience changes, right? So from having the customer, they didn't know exactly what [00:24:00] they're searching for to helping the customers start with their problem and their context and intent is a very different way of serving the customer's intent in that sense.

One piece I just want to kind of dig into here is like thinking through, you know, we were just talking about there where ostensibly the person could be, you know, it goes from, right, like I'm looking, I do look great in a red party dress, I'll be honest, right? But you know, the difference being rather than, you know, maybe I have a retailer or two who I.

General and loyal to or or just used to shopping at. Historically, my move would've been go to their site, look through and see what red party dresses they have. Yes, I'd probably pick one of those. But now does it start to open up maybe a world where more retailers can get different share than they used to?

Because now I can just say, Hey, like what Gemini now has person data so it knows my calendar and all that kind of stuff. I can just say, Hey. I need an outfit to where did this event coming up on my calendar? Right? And it can go do the, you know, shopping deep dive and come back and say, here, looking at that and knowing your preferences from past shopping and what you like, what you don't like, here's five options I put together.

And it may be retail agnostic, it might be an outfit [00:25:00] from like three different retailers. What does that mean for kind of established retailers versus up and coming? Is there a chance for people who maybe aren't as big to get a shot in the arm of growth? No, absolutely. I think it plays very well for both.

And to me it actually has. Filter the balance a little bit in terms of the retailers who were difficult to discover otherwise in a very traditional Google search. Right? So for example, in retailers with a long tail of products, right, which were difficult to discover otherwise that, because those products would not get a lot of traffic.

But in an an agent search world, because you're capturing the customer's intent and because of the way those agents would actually go and research some of the sources, not just look at. Keywords and not just look at traffic, data, et cetera, allows you to now surface products that otherwise would have been difficult to discover.

Right? So lc, for example, in terms of what they're doing or any other retailer with a very long tail, difficult to discover now, I think has a better chance of getting discovered in that surface scenario. Let's go to the data side. Like, you know, are there things I need to be doing [00:26:00] on a data setup or data cleanliness, both to enable myself to do as a retailer, the smart personalization that, but also to, to take advantage of this move in ai.

I assume there's things I can do wrong where I'm not going to be discovered in it. And then there's also things I need to do on the data personalization side for people who are on my site or maybe even interfacing through an MCP server or some kind of. Commerce flow. Exactly. Now. Now my perspective, there's an and not a new perspective.

Some of the people have also sail it, but from my experience, AI needs actually containers before it needs any models. And what I mean by that is, let me give an example. So global trade did not scale because ships got faster. You know, it's scale because everybody agreed to the same context or the same standard, which is the shipping container.

So the one standard box. Really made handouts predictable across ports, across trucks, across rails, across insurance companies who could now standardize their insurance rate. Mm-hmm. Because they could now know the risk across freight companies who could now standardize their rates because they exactly know what they were [00:27:00] cutting and the volumetric rate of that.

What increased really was the reliability enabled in just in time systems and risk management, and the power really shifted to orchestrators who engineer the whole network. Right. Yes. Standardization of the box made it possible, but everything, the whole system. Then sort of reinvented itself around that box.

So the ports, the trucks, the rails, the shipping containers, the ships themselves, the sizes of them, the cranes on the ports, the automation of that, everything was working around that box. So to me, I'm using the same analogy, which is AI needs the same thing, which is containerization for data. Now, if your product data, your customer data, your event data, your supply chain data, such in different systems, or CMSA, C-R-M-A-C-D-P, or RM.

You don't have an AI strategy, you just have a translation debt. What we need to therefore create or start with in this context, therefore, is to create these dependable boxes, which is chemo, ontologies, and APIs, where AI could infer them from messy inputs and really have the same [00:28:00] standard input for every system.

It is picture, whether they are onsite or offsite at that level of strangulation of data is really what is gonna make it possible for retailers, both large and small. To really play well. And within that AI world, how important is it as a maybe smaller retailer for me to figure out how to best box up and containerize my data about the product versus just having it available on the site somewhat intelligently.

Just having it there. It seems like realistically an AI should probably be able to go understand the data there, whether or not it's perfectly formatted the same across both those things. Right? So may not show exactly similar on different surfaces, but I'm talking more about the underlying data, not necessarily how it really shows up.

So that's one. Second is it really matters that you have standardized, continuous because other agents were gonna search for your resources, they're gonna make their decisions about whether. Your content or your data or your product is relevant to the customer's intent in sub milliseconds. If you're gonna have a messy data, you just gonna add [00:29:00] latency to that AI agent's work and therefore increase the chances that your product may not get picked up or surface in that particular search for that particular intent.

So when I may send some data, I just don't mean in terms of having the right data attributes. So let's take for example, a product container for an e-commerce, right? So you would have containers typically even today, like attributes of size and color. Or you could have attributes around protein time, which is price, your availability, your delivery, ETA, or you could have signals like reviews and customer ratings and returns.

But more importantly, what you need to add to those today is relationships. Which is, this part is similar to some other product or this part is best use for this product, or this product has the best use case or customers find it most useful in this particular weather or this particular context, et cetera.

And add some concerns around that. For example, saying, Hey, this is the top 20 to 30%. Product categories where I can add some sort of intent, which is, for example, saying, you know, back to school, laptop under [00:30:00] $400 for a 10 to 13-year-old. At the moment, I start to layer relationships between my products, not just the attributes and also some level of intent.

What that does is it makes it possible for my product, therefore, to be able to be searchable or being able to be discovered. In an agent search, much faster, much better. Because I'm now talking to the intent of the customer, not just the attributes. I was doing. Search in a keyword search era. Attributes matter, but I'm trying to match the customer's intent.

I need to go beyond just the attributes. I ask about that because I have a background in marketing and one of the things that always really bothered me about Google and search in general and kind of how we discovered things in the world that we've lived in for, you know, couple decades at this point, was, sure search could be a great democratizer, right?

But in reality, a lot of small businesses had great product. Something that end users could really, really benefit from. But they weren't big enough to have a search person. And as search got more and more and more complex, it started to be you needed all these [00:31:00] technical structures on your site or you need to do certain things.

And you think about like the small business where maybe one or two clothing designers, uh, a production person, like a five person company, there's just no world you were losing because. You had great product, but just you weren't formatted the correct technical way for Google. Yes. And it seems like this is a chance to kind of solve that problem and give that small business a chance to compete at the same level as an Amazon because the AI can parse that data without it having to be magically structured in the exact way the robot requires.

But I guess the counterpoint to that is at some level, these are still for-profit companies that are running the ais. They want to spend as little money giving you the right answer as possible. And so if you have it structured in the right way and they can discover it easier, right? To your point, the latency is lower.

They have to spend less token, right? And less compute parsing the data. Maybe they start to lean that way anyway, and it kinda erases that advantage. But it seems like that'd be a nice world. Exactly. This is not just the same as a CO where you have a certain structure on your site and you sort of, you know, add the right words, et cetera.

It's really to me about whether you understand the [00:32:00] context that your product is gonna be used. Mm-hmm. You understand the customer segment or the customer type that's gonna prefer your product, and whether you can match what you're putting that out in terms of the content or whatever to match the intent.

Right. The, the more you do that, to me that really matters more and the sort of standardization of data and relationships of products, information, et cetera. More than it being structured in a particular way just because it suits and how an engine is gonna discover you. So to me it actually benefits, I would say even smaller retailers or retailers with long tail that wouldn't be able to pump in as much money in terms of standardizing their issue, practices, et cetera.

To me, they sort of have a level in an agent search. Uh, but again, we'll learn how that goes to your point, right. Like you just said, we'll learn how it goes. We're not there yet, so we're gonna see. We'll have to see how it goes. Yeah. But I think you hit good points of it's evolving quickly. Don't kind of tie into one tool too fast rate, focus on the problem, not the tooling.

And then, you know, hot swap, the best tooling for it. And I like the idea that you had of. Capture how people are going to use it, capture those [00:33:00] relationships. 'cause that's something essentially if, if the LLM is just able to parse and read write, you don't need to do all the technical structuring and get into like headers and metadata and stuff like that.

If you can just tell the story. Well, it should be able to interpret that and like opens up a new world. But yeah, to your point, we're gonna see. But you know, Rahul, as much as I could go on talking about the future of e-commerce now for hours, I'm gonna be completely honest and it has been fantastic. If people wanna reach out and kind of ask you more or just get in touch, is LinkedIn the best place to reach you?

Yeah, absolutely. Ping me on LinkedIn. I'm happy to answer any questions. I post there fairly regularly, not as frequent as I would want to, but yes, I'm fairly available on LinkedIn. Again, Jeff, thank you so much for having me and uh, this was a very enjoyable conversation. Great chat, and Rahul, let's stay in touch and hope to have you on again.

Thank you. Bye.