Transcript: Agentic Commerce & CRO: Winning When Your Next Customer Is an AI — Itai Horn (Resident)

Eyal David: [00:00] Hey, this is Eyal and you're on Product Builder We're about to start the episode, but first give us a follow on Spotify or subscribe on YouTube and help us reach more ears. Alright, let's get started. Itai, what's up?

Itai Horn: [00:20] How's it going, I'm here with Eyal, great to be here. Great that you're back — Itai is back with us for another episode. This time we'll also turn to

Eyal David: [00:28] really, for anyone who wants to hear a bit more about Itai — it was an excellent episode, one of the first episodes on the podcast, we still get feedback on it We were just talking about this before recording. So basically Itai comes from Resident, he's a Senior PM, and basically — tell us a bit a bit about Resident, it's really interesting. Let's say we came to talk in this episode about the day-to-day side and the forward-looking side, the agentic side of commerce. You'll give lots of tips here that we'll save too — examples, your best practices for A/B testing and e-commerce. But let's start — tell us a bit, maybe tell us a little about Resident,

Itai Horn: [01:04] what it looks like? Gladly and happily. First of all, thanks for having me and it's great to be here. So basically I'm a Senior Product Manager at Resident, I've been there for over two years. What is Resident? Basically Resident is a US leader in the field of e-commerce for mattresses, it's called DTC, Direct to Consumers. It was actually acquired by Ashley Home about a year and a half, two years ago, for over a billion dollars. Ashley Home is in furniture and also in mattresses in the US; it's very strong on the physical retail side, it also has a website and brands, but it's more of a leader on the retail side, and it acquired Resident in order to add the power of e-commerce, of online, of digital marketing and the whole world of conversion and optimization that we'll talk a lot about today. So, you told me a bit about the industry — blow our minds with some numbers. The mattress industry, who knew? The mattress industry is about $17 billion in the US, and it's an industry that has a lot of room to grow, and we're very focused on how to generate, to increase the conversion rate, to increase revenue, conversion rate, gross profit. It's really interesting how the business works

Eyal David: [02:23] but it's interesting, because it's a product you buy physically by definition, you need to test it, right? And you're basically moving that online and you're talking about crazy numbers, right, and we talked about this just before recording, that it's basically like a few events in life, right? Maybe tell us about that for a second — buying a mattress, right? Yeah, yeah, so there aren't too many events in life

Itai Horn: [02:44] it's basically an event, like when you move apartments, or when you grow up as a kid, or when you want a bigger mattress because you moved, upgraded your apartment, bought for the kids — and these events, so this event is one that involves a lot of research. The users who are going to buy mattresses do a lot of research on Google, on Facebook, TikTok and other digital marketing, and today a lot is also coming in — you're talking about this, the LLMs — users who go into ChatGPT, go into Gemini, Perplexity and other places, basically to do this research and figure out which mattress is best for them.

Eyal David: [03:26] In words — to even measure it. I picture myself going to the store and lying on all of them. Okay, so as we said, so in the end you'll really give tips that actually worked for you in e-commerce, and you'll share them, so come on, tell us a bit more — so how do you actually measure success at Resident? Yeah, so we're basically — it's very, very data-driven.

Itai Horn: [03:48] The role I do, and at Resident in general, we have a big data department with data scientists, with analysts, and basically every test we do — we do tons of A/B testing, every test we put live is data-driven, we look at several metrics, the main metrics are — we try to increase revenue, we try hard to increase the conversion rate, to increase gross profit, our profits, and we look at one KPI that's very specific to the test we're doing. So product mix is basically a KPI where the more we drive users toward the more expensive mattress or product, the more we increase both revenue and gross profit. and it's mainly — it's a very significant KPI we look at, derived from all kinds of tactics we use, tactics related to user experience, tactics related to the A/B tests we run, tons of tests we've run over the years, and data that we have internally as well as industry data and studies we read, related to all kinds of research bodies.

Eyal David: [05:04] Super interesting — when you talked I thought, hey, you didn't mention LTV, and we talked about it in the context of events. But tell me, for this industry — I currently run a short-term rental apartment, and I think about it like this: there's basically also a rise in the short-term rental market like Airbnb and such, so suddenly the rise in online mattress purchases became significant, and maybe the LTV of a single user suddenly skyrockets. Is that something you see, like, is there a correlation between these things?

Itai Horn: [05:34] Yes, we actually see that — the more open the market becomes and the more users are adapted to buying things online, then even these things that require more research they'll end up buying, and in the US this has become very, very strong. You can buy, let's say, a mattress and an adjustable bed, quite adjustable, for over two thousand dollars, and people do this a lot. And they also have the option to return within up to 365 days, so that's also something that provides certainty — the option to return during the year.

Eyal David: [06:11] It really is a matter of trust, as you say. If we look at challenges — you just mentioned trust challenges, that's the first thing: can I buy sight unseen, like what I see, what else.

Itai Horn: [06:24] I think the brand too, in the end — you're in an industry that's very brand-driven, and Resident worked very, very hard to create several brands, brands that are now leaders in this online mattress space. There are two, they're very well known in the US — Nectar and DreamCloud — brands that the moment you're a user in the US and search on Google, Facebook or one of those, they'll show up first. Um, mattress, adjustable frame, which are the strong products, and other products, and once you build this brand awareness, and the users who come and buy and know the brand, and a site that's very strong UX-wise, a site that knows how to give you the product recommendations in the right way — then they'll come and buy and complete the purchase, and afterward too, even after a long time you see there are very good reviews, which also really helps the purchase and the whole process, so yes.

Eyal David: [07:28] So no, that's great, but let's actually talk about challenges — but what you said is interesting. I basically picture you as a store that sells many, many products, since you mentioned mix. So basically traffic that comes from all these different sources lands on the homepage — that's an e-commerce problem in general, right? How do you handle its navigation now — it's a very tactical thing I'm asking — how do you handle navigating it to the right place now.

Itai Horn: [07:56] Yeah, so usually in e-commerce you land them on the PDP, the product description. You mentioned strong brands, as I said.

Eyal David: [08:03] Your brand is strong — they land on your homepage, right? Not the homepage; most of the traffic lands on the product page, actually from Google, Facebook and other sources, and from there you basically want to run all kinds of tactics.

Itai Horn: [08:16] UX — through the tests you build, together with the user, the right way to work, to lead them to the golden path of e-commerce. The golden path is really that they do the research they need, and as fast as possible buy what they want, while they go through and do the research. Whether they come with intent or without intent, you basically try to add more products they need, upsells — whether it's a mattress, another bed, add accessories, in order to identify the user's level of readiness. So it really depends on the level of readiness, or as we call it, intent — so it really depends on what it is, a bit on the source they come from, whether it's Google, Facebook or other places. To identify their intent level or readiness level, and from there we know what we need to do to convert them, to convert that user.

Eyal David: [09:10] So what does it do, tell us — you promised some tips. Yeah, gladly. So we've basically run tons of tests in these areas,

Itai Horn: [09:17] across the different brands. I can give a few things that you can already see on our sites that work well, after we ran them for a while and they really delivered results on the KPIs. So one test was really to show the... there's a payment provider called Affirm, and it basically lets the user pay not all at once, but to pay monthly, to pay in installments. We're very familiar with this in Israel, but in the US it took a while for users to start understanding it, and today it's also become very common in the US. Basically we put a component under the add to cart, under the 'add to cart', for the payment, that basically shows the user how much it would cost to pay for it one time, versus how much it would cost to pay for it monthly, and that illustrates to the user — you see $100 a month versus, say, $2,000 total, and that really shows the user it's not so scary, and it's actually not that expensive — you can pay monthly and it's less of a burden on your wallet. This worked well on pages with a high AOV, a high average order value, like your average cart is high, high. where you can assume that people with high intent land and come to complete, so you come and help them make a decision even faster. Absolutely, you come and help them make a decision faster. We saw that it also raised the conversion rate, it raised the revenue very, very nicely, and raised the product mix, which basically means we led the users to buy the more expensive products.

Eyal David: [10:53] Yeah, because you're basically telling them they can buy more and pay for it over time. Right. Okay, great. But, in terms of context, as we discussed — when I go over it, if you have nothing to add — but about the transition from the homepage, or the top of the funnel, to the desired place — you ran tests there too, right? Yes. So we ran a few tests,

Itai Horn: [11:15] we have this test on the homepage called dual CTA, which is basically what we enable for the user, once you land users on the homepage, or land them on any page in general, they may come with different intent, users who came, say, from Facebook, and users who came from Google, or users who came from another traffic source — and actually not just the traffic source, but what they're looking for — whether they're looking to buy a bundle, whether they're looking to buy a mattress, or a specific product, or whether they're looking to get a bit more information. And this test, we basically ran it, the goal was basically to address several types of users within the homepage, and in the end, when you land today on the homepage of Nectar or DreamCloud, you can basically see two CTAs, two calls to action, two buttons that lead you to the next page, to a product page. It worked for us — we basically saw that users do want another place to move to, not just to a regular product page, but to the bundle page. Okay, so you added another option for them and it worked well, I see.

Eyal David: [12:26] Yes. Just before we continue — click the link to the Product Builder WhatsApp group in the episode description, and join the community. There you can continue the discussion, ask questions and be the first to get updates on new Tachles episodes. We're waiting for you there. And now, back to the episode. So we talked about intent, and if I now arrive from an LLM, then suddenly there's — we'll talk about this more in depth — but it's really an experience where, okay, you don't know with how much intent or how much knowledge I arrived from the journey I went through before I even set foot in your store, right?

Itai Horn: [12:58] Yes. So basically this bundle place — you're saying, okay, I slow down the decision speed, and I let them enter something more relaxed, that's what you do. Right. It's actually a very central page, because the AOV there is very high, and you want as much as possible for the user to have as much information as possible, whether it's visual information, whether it's copy that's strong and leads you to purchase — that's the way to convert them into a paying user and lead them to the golden path.

Eyal David: [13:23] Great. Okay, are there more best practices you want to share about experiments you ran before we move on? Just to say that we run tons of experiments — it's our bread and butter,

Itai Horn: [13:34] like, to succeed you need to run a lot of experiments, have a lot of data, and also use a lot of... a lot of techniques, to know when and how long to run each experiment, how much traffic you need to bring in — there's a difference between the different pages, whether it's a bundle page or, let's say, a PDP or homepage, it'll take a different amount of time and you can expect different results, and additionally to know when to stop an experiment or drop some of the variants, sometimes you run more than one variant — say you run three or four variants, so maybe you kill two of the variants to reach a decision faster, so it's really a whole world, it's a super interesting field, it's a field of succeeding in companies that are performance-based, and they try to increase the... And you're a company with really a lot of traffic, so maybe let's think for a moment from the perspective of folks who are starting out,

Eyal David: [14:28] or folks who don't have enough traffic — so maybe give some insights on how you brought AI into your workflow, as I said, as we discussed, it seems very relevant to me also for entrepreneurs at the beginning of their journey, or not at the beginning — maybe share a bit about that.

Itai Horn: [14:42] Yeah, sure, sure, gladly. So we're basically also working a lot on this area of agents, and a lot on, basically, turning us and turning the company into AI-Powered Product Managers, or AI-Powered Employees. Which basically means making better decisions and making fewer mistakes in the end. Both making better decisions and making fewer mistakes, also doing more tests, also working in a more, let's say, informed way, and also becoming more efficient, that's the main goal. We do this through... I can talk about this side too, the agents we built, and how we work with them, and also the side of what we'll surely talk about — Agentic E-Commerce.

Eyal David: [15:31] Let's start, tell us what you did. Yeah, right, gladly. So we basically work very, very hard with AI,

Itai Horn: [15:39] both in terms of getting suggestions from the AI — how to build the PRD in the most correct way, how to approach design in the best way, how to use all kinds of vibe-coding tools to build better designs, that will succeed in A/B tests and bring us results. Maybe tell us — it's really interesting — what are the inputs for these things?

Eyal David: [16:04] Yeah. What are they? What do they get? What do they run on? Sure, sure. So I'll describe something very specific — besides the part of the AI that supports us,

Itai Horn: [16:12] the agents we built, into which we basically put all the techniques we know, data — we actually call it the CRO Agent, which is Conversion Rate Optimization, an agent that's supposed to help us increase the conversion rate, and it's based on both our data and all kinds of data that ran from the industry, data that comes from Baymard, a research body that presents a great many tests that huge retailers ran, and other inputs we put in on the context side — we basically gave it context on what to test and how to test, and it basically helps us create automation in this whole process of ideation and execution, in order to generate more tests, to generate tests that are stronger and winning, and it's very evident in the results. So — let me just stop you,

Eyal David: [17:05] so if I wanted to start building something like this now, and I don't have all the accumulated knowledge you have, I could really bring in a lot of knowledge from outside, about tests that worked, didn't work, user psychology, what works in commerce, e-commerce, right, and maybe even all kinds of books and studies, and then I'd already start from a much better place, so to speak. Absolutely, absolutely, so it's like if you really take it to the place of, say you start from scratch — context of my store, my audience, etc. Yes, my audience, the conversions probably, traffic sources, types of ads, the site itself that does a real review, the agents really come and do — our agent comes and does a review of the sites themselves, and basically data on which tests succeeded, which tests succeeded less, on which pages we ran them, and what really made them succeed, and in the end, this whole thing really helped us increase the velocity of the tests we run, from 2 to 4 tests running simultaneously, to 8 tests simultaneously, and also the success rate, from 10–20% to 35%, and this whole process too — today I can come to this agent and say, go generate me a PRD, help me strengthen the hypothesis, help me generate all kinds of prototypes, mockups that I can present within the company, both to R&D and to Leadership, and give me all kinds of suggestions, on how to do the A/B tests in the best and most informed way. So on the face of it, you're saying this helped me generate more tests, but if you have less traffic, it can even help you generate fewer tests, but more precise ones. Right, it makes them more precise, it's basically another team member that sharpens us and helped us in the process. And how long did it take you until this thing became operational? from the moment you started running it until... I think a few weeks, roughly, and we keep improving it all the time, and each time we work on it together and try to improve, and that's how we expand its context, and the explanations of what works right — it's really refining on the fly. Great, and you also told me another thing that's unrelated to this, Yes. But it is related to the number of tests you need to run. You have a few cardinal areas in the store where you must always run tests for sanity, right? what does that look like? Yes. So basically, not just for sanity, but for... So areas we're very focused on for optimization are areas where the money is — whether it's the product page, the PDP, whether it's the bundle page which is super important, whether it's checkout — there are a great many tests that are definitely run in the industry, not just us but other players too, and whether it's other places like the homepage, navigation, tests that run across the whole site, cross-site, it's called. Yeah. Okay, so we're talking a lot — okay, and you mentioned that you also really work on prototypes, it helps with that too — I hope I didn't miss anything with what you mentioned. We talked a lot about basically the experience of how a user today meets the store, and how we can optimize it — and how, alternatively, it looks from the agentic side? Yeah, great, so it's a field that's become very popular, also a field that's become super, everyone's been talking about it lately, and we're getting into it too. Now, I don't understand why — because half the traffic, or I don't know how much of the traffic, stopped arriving, and suddenly it's attributed to a different source, right? Yeah, it's attributed to a different source, and if you're basically not there, you'll lose the traffic, you'll lose your positions, you'll no longer be a leader — you have to be there in agentic commerce, a little explanation of this field of agentic commerce, from my side and how I see it — so we basically have a few types of agentic, first there's the Nanagentic, which is basically the classic model that we work with today, it's basically User as an Operator — User as an Operator, comes from any traffic source, whether Google, Facebook, other traffic sources, arrives on the site, no matter which page, clicks add-to-cart, reaches checkout and buys — that's the Nanagentic, the Traditional. Then there's the Semi-Agentic — that's a user who basically goes into one of the LLMs, ChatGPT, Gemini, Perplexity, Claude, really one of the LLMs, and tells the LLM, help me, find what I want — say, for example, help me find a mattress, between one thousand and fifteen hundred dollars, on the best sites in the US, and it basically leads me — I get a kind of comparison, a comparison table, within ChatGPT or within Gemini, and from there I move to the site. That's the Semi-Agentic. You move to the site, there are a few cards and options, and then you move over? Yes. And soon the whole world of advertising on OpenAI will enter too, right — we can talk about that as well. Sure. But the experience still happens within the vendor, the buying experience itself. Right, right. The experience happens within the site, within the Direct to Customer, the e-commerce — which is where a great many, let's say, players are getting into right now, and many players are already there — and the third thing, and this is the most ideal thing, let's say it's the Full Agentic, what retailers and e-commerce companies really want to be — it's the full experience, basically you build an agent in ChatGPT or in Gemini, or in any other LLM, and you tell it, buy me this mattress, for a thousand dollars, on one of the sites, or you give it freedom, and it — that same agent goes and does the research, and makes the purchase, from the moment you give it the... you're okay, it does the whole funnel for you, and it's relatively new — it's the last few months, let's say the last half year it's entered these worlds, all the players want to be there, and I can tell you there are all kinds of very large players in the US, Etsy for example — it's a big marketplace in the US — that's already there and works with Stripe, which is also a giant in the payments space, and offers the full agentic in certain places, and is testing it right now, testing these places, and these are the places that e-commerce and direct-to-customer players ultimately want to be in. I'm not sure, you know — I'm thinking out loud. First of all, the Etsy example is good, because okay, they dominate the crafts market or something like that, right? But if I'm now, say, if I'm Nike, I think, and I have, I sell, but lots of other stores also sell my merchandise — I want them to buy from me, I don't want them sending me off now to some mom-and-pop shop. So my way to stand out is basically only price in that case, so I don't know if that's what I'd want. So let me tell you something that e-commerce players are actually doing, and what they're putting a lot of effort into right now, is being first in these places too. The more you're in the citations, first within the LLMs — just as you now want to be first, that competition to be first on Google, first on Facebook, first in all the traditional places — likewise you want to be first and leading in these LLM places. And the more you lead there, the much easier it'll be for you to win in these places too. Now, I really connect with what you're saying, in the context of Etsy, and it connects to this world where there's both risk and all kinds of challenges. I heard a podcast recently with the woman who leads — the Chief AI of Stripe — who basically says, she works with the biggest retailers — we also work with them, with Stripe — and she says there are a few points within agentic commerce. First, control — the biggest brands in the world want control in their hands. Once they control the LLMs too, it'll preserve the control. The second thing is discoverability — finding their brands, or as you touched on, they want to be first, they want to be not only on Google and Facebook, but also first in the LLMs. The third thing is fraud. What's fraud? like, until now there were tons of bots that would come in and we'd block them from commerce sites. Now you want to let these bots in, you want to label the good bots and the bad bots. That's the third thing. And the fourth thing touches on, it's basically checkout. How do you create an experience that strongly supports checkout, and also leads the user — one who's in full agentic, an agentic user — to reach the golden path, and make purchases, and complete the... It's funny that she of all people says this, because Stripe took the checkout experience, they standardized it, and they did, you know, one checkout for all, and so on. Yeah. So it's funny she says that. Yeah. You're right, I think if we think about it for a second, it's basically a kind of long-term, short-term, okay? At first, as a business, I want first of all to be discovered, and to appear, first of all. Yeah. Then comes the more distant part of it — yeah, I also want to control the experience, if we take Resident now and flood it with — I don't know who the players are — Bed Bath & Beyond, and IKEA, just saying. Yeah. Then it'll be purely a price decision, and suddenly the brand doesn't matter much, or it'll be very hard for new players to build a brand here at all. So I want to control that, right? Yeah. But how do I manage to get the user out into my experience? Right? That'll basically be the challenge here. How do I get the user into my experience, but when they're already inside the LLM's experience? Also how am I first, how do I... all the positioning, I'd say, that's the short term, I want to appear first, I want to be seen, and to appear first. Yeah. Then comes the game of, okay, I'm first, or I'm second, I'm inside that zone of influence you mentioned, how do I pull them over to me now — and here it's also interesting because all kinds of other things come in that you can tie in here, like experiences that only I can offer — I don't know how it is with mattresses but say with shoes, there are already players offering the try-it-on — Shopify just announced it a while ago, so now maybe with Nano Banana it's easier, because until now it was in AR, right — the app would place the shoe on you and you'd see how it looks so that's a reason to come to me, since they probably won't offer it — yeah, yeah, so I think it's a good example, whoever offers try-it-on it's an example that's a bit less strong in the mattress world because there are some challenges in those areas so in the mattress world — I imagine, I'm speculating — it could be that if I went to Nano Banana I'd manage to take this mattress and place it in your room, and that's an experience I'd offer and they wouldn't — for the sake of it, yeah, how does it look in your room? IKEA already does this in Israel; in the US there's some data on it that you're familiar with — they probably keep doing it it probably works for them in A/B testing. It's interesting, it's something interesting, but you have to do it right — until now it was always hard to do it well, like to give you a good enough experience. I think today too, with the improvements, say, in the capabilities of mobile — because we're really very mobile-focused — it can work. Yeah, we can talk about how IKEA does this against the strategy they'd followed until now — that you go to the store, you see their landscape, right — in the back. So no, they're saying I'll now place it in your home so that I stand out — yeah, and you'll see how it's probably not just A/B testing, it's simply a strategy they're pursuing yeah, a strategy they're pursuing now, I think both in the context of LLMs and in these direct-to-customer worlds in general mattresses and these worlds in general — you really, really need to be focused on how you explain the offering to the user — it's one of the biggest studies running at Baymard, which is a large research body, and also in the internal data we worked on — the better you explain the offering and the proposition to the user, whether through visuals, through copy, right, through keeping it simple — that's how you lead the user into the golden path, which is ultimately important. But it becomes much more challenging when you really only have a card — I assume that's how it'll be, right? Yeah, when you only have a card, like you're saying. There's a card, like you have the product image, you have a price and you have some title, right? Like you're saying you see it in ChatGPT just the card — yeah, it becomes — but if it's full agent, then you expect the agent to go into the site and pull all the information from there, and then — like, you'll feed it enough information so that it executes, now you tell me, and today in such organizations, who's the person responsible for this today? I mean, it's not the uncle who does SEO, because he already understands the problem, right? It's not him. Yeah. He probably tried to index it. Super interesting question — broadly, actually someone who does come from these worlds — it's roughly the uncle who does SEO, but that uncle who does SEO needs to make a transition to LLM, like to become an expert in LLM — so we have a few people who deal with this and really get into these worlds, how to be discoverable to LLMs, how to be first, how to be in all kinds — there are tools that measure ChatGPT and Gemini, what your discoverability is, what your rating is, where you stand relative to... Exactly, it's suddenly a specific product and a specific person. Yeah. So a role is born here, it sounds like. Right, it's an expertise of LLM Search that's emerging on the fly, whoever does it really specializes in it, works closely with R&D, understands the methods, and there are tools you can use to understand where you stand against the other players, and to see — just as you see SEO tools and SEM tools, there are also LLM tools to see the ranking. Which is kind of crazy until now we always aimed for the 80-20, right? Yeah. In these areas now, no — now it goes very much to the long tail, the picture needs to be very precise, and you need to see — it's like that, yeah, it's like that, now something else we didn't touch on but maybe worth raising — in the end you also need to be an expert in this, but you also ultimately need to understand UX tactics, marketing and e-commerce tactics — we really love to use — it works very well for us — the Good, Better, Best tactic, which basically supports the KPI of the... you go back to human — okay, we sent you to land from the LLM onto Itai's site, please — and I believe the LLM too, in the end, will go there, ultimately looking at all kinds of ways you present your offer in the best way for the user, and it's one of the tactics that... Do you want an example around this? Yeah, broadly, Good, Better, Best — you can also see

Itai Horn: [32:01] Netflix does this and... let's say other sites do this, it's either packages or tiers or products by cost — basically the Good, Better, Best at our place is called Classic, Premium and Lux, it could be any other name, but basically it's price tiers of the same product, and you want to show your offer such that the more expensive the product you buy, the more you get — and our site is very, very oriented toward these worlds of Good, Better, Best — you can see it on any site, it's not just us, other players do it in the process too, and I believe the LLMs too have gone to these places where they look at these techniques, look at who does the offer in the best way, and will also rank them for it. That's exactly it — if we talked about 80-20,

Eyal David: [32:49] you look at the packages, for example. Yeah. So suddenly it becomes per-specific-user and their needs, right?

Itai Horn: [32:57] Yeah. So it could be that the package itself updates according to the customer's need and where they live and the delivery time — it could be that it'd be cheaper to ship, just saying, to bring it now to someone from Chicago than elsewhere, right? Absolutely, absolutely, you're touching on it,

Eyal David: [33:11] yeah, super interesting point — personalization basically becomes much smarter, like, together with the LLM you can create much smarter personalization in these contexts. So super interesting, we'll really see where this goes are there other things you can share about what you see from your day-to-day work that you encounter there?

Itai Horn: [33:32] so I think it continues to be tons of data, tons of data — on my side, if we touch for a moment on AI and A/B testing, it's to try as many tools as possible, like vibe coding, whether it's Gemini, Google AI Studio, V0 which I really love working with, and other tools — Claude of course, we didn't mention its name, but it runs everywhere, and Claude really helps our developers too, and we also use Claude for optimization and understanding, and... And you do optimization? So also also all kinds of things related to giving recommendations on prototyping, basically to help us with this whole process — agents we built in GPT or in Gemini, I also put them into Claude to examine them, give me a rating on them, and maybe improve them and give all kinds of recommendations in these areas

Eyal David: [34:30] yeah, funny — the guest who was here told that when he started working, I don't remember which tool he uses, but he interviewed them all yeah, he gave them one month each, he does this — why should I keep working with you and not with someone else — so that's how he basically tell me — and basically in prototyping are you already pushing code to production? No, right? We don't do that — the big challenge is basically connecting the design system

Itai Horn: [34:58] into the vibe-coding tool — everyone's working on this now, everyone's working on this, there are a few startups working on it — no, I mean companies themselves now everyone's setting up the design system — things they'd postponed — it also requires all kinds of, like a design system block, and also this part of connecting it fully so it works smoothly there's still friction there, but yeah, it really helps us with this whole design process — we call it design review like creating prototypes quickly based on our Figmas, presenting them, working together with the designer, presenting them to R&D, to Leadership, shipping to production faster with probably fewer bugs

Eyal David: [35:40] Great, okay Itai, we're reaching the end — really interesting. Tell us, what things do you do besides work — what does it look like, what else do you do, how do you contribute to the ecosystem? Cool. So I really, really love

Itai Horn: [35:53] the product management ecosystem — over 12 years, over 12 years in this field recently, in the last few months, I also joined the product management program of Google N.A.I, Google, Google N.A.I, Reichman — and basically in this program I get to do mentorship for people who are entering this world, who want to get into this product world — I help them with AI and with other tools, and that way I'm very much inside the industry, living the industry

Eyal David: [36:33] meetups, mentorship. Great. Okay, so — you've got a lot of words here, that's good, it'll get thrown into our queue for transcription and we'll index it — I'm joking. But tell us now, when young folks come to you wanting to get into product, when you don't even know where this profession is headed, the whole profession, right? How do you get into a world that's so hard

Itai Horn: [36:54] to break into? Yeah, so interesting question, also because there's this talk now about whether to even learn programming at all, the whole thing of being an engineer or going back to the old world — I think what I mainly emphasize, so what I do is basically: try as many tools as possible, work with as many tools as possible, try to see the limitations of each tool, for everything you do, do it first with AI — use GPT, build agents, use Perplexity to do competitor research or research in general — use AI tools as much as possible meaning as hands-on as possible? As hands-on as possible — it's really become a world today, it's like I heard posts here, it's going around a lot on LinkedIn, that the CEO of Spotify, like, part of the... sorry, Shopify — the CEO of Shopify is actually heavily involved with Claude Code, and gives directives, does prompt engineering for Claude — so it becomes really a world where even the C-level and everyone gets back into being hands-on, wanting to feel it with their hands, suddenly developing again people who weren't in programming are back and are writing all kinds of things in Claude Code... I'm really glad you said two things, first of all

Eyal David: [38:11] we really didn't talk about Tobi and Shopify — this whole thing we talked about with the agentic and basically the MCP Apps that we didn't mention — so yeah, Shopify is already going there, so it's maybe another stamp of approval on everything you said. I don't know if it's about getting hands-on like it's the desire — it's simply a home crowd, everyone's scared, right? So everyone wants to be there because no one knows where this is going

Itai Horn: [38:32] yeah, there's serious FOMO in these worlds, let's say — my point, I'll tell you, is that this podcast, I started teaching it to product people and it changed a bit over time — it's not just the audience that listens, really not — but my thinking is basically that it's now product and entrepreneurship, and also that product people, logically, go on to become entrepreneurs right after — that's also something I see

Eyal David: [38:54] so that connects exactly to your brand, Product Builder — in the end they become builders it's talked about a lot — it's not just engineer or designer or product, but builder right, they combine — we have a mutual friend, we talked about this with Harmony — they really take all the frontend and all the backend and teach them — they consolidated the guilds because there are more people, double, to go over more code, so it makes sense I think that's good advice — go build as much as possible, that's basically what you tell them, get your hands dirty in the end it's about building businesses, that's what we do — so good advice all in all. Great, Itai, so great that you came back to be a guest and I remember people reached out to you a lot on LinkedIn, so anyone who wants — I'll gladly say people can reach out to you on LinkedIn, right? Gladly, happily, yeah so alright, thank you very much and until next time — thank you, I'd be glad to be here again soon — thanks, alright, that's a wrap bye bye, bye bye. Hey friends, thanks for listening — if you found this podcast valuable you can subscribe, follow us of course for more episodes on Spotify, Apple Podcasts or any other app and of course, if you didn't find us on some app, I'd be glad if you wrote to us — we'd really appreciate five stars on any platform and that you follow us so more listeners can discover us and find the podcast — you can also find the previous episodes on any app or on the YouTube channel — we have links in the description. So until next time, alright, be efficient and bye bye