Transcript: Zeevi Arnovitz | Wix | How Are AI Tools Changing Product Management?

Opening and meeting Zeevi

Eyal David: [00:00] Hey everyone, it's Eyal. You've reached Product Builder, where you can listen to the most interesting product conversations. My journey started in 2012, when I learned what product is at my own startup. Since then I've worked in the industry, and since 2017 my company has been advising a wide range of companies. My drive is to help all of us get better as product people. And before we start, if you enjoy listening to Product Builder, please rate us five stars on Spotify. It gives me great feedback that the podcast brings you value, and it would make me very happy. So let's get started. Zeevi, how's it going? Good, Eyal, thanks for inviting me, excited to be here. So glad you came. Tell us about yourself, you're product at Wix, tell us in a few words about your path, who you are.

Zeevi Arnovitz: [00:50] Sure. So I'm Zeevi, I live in Tel Aviv, married to Tamar, and today I've been a product manager at Wix for three years. Let me quickly tell you, broadly, how I got into product at all. I have no technological background, not from my studies. In the army I was a combat soldier, and when I got out I realized I wanted to get into high-tech. But I didn't really know it; I knew there were developers, that was basically the role. I did some quick Python course and said, okay, interesting, but let me see what a company looks like from the inside, what roles even exist. I got accepted to a company called Movao for a role that was literally checking spelling mistakes, that's basically what I did for three months. I checked spelling mistakes at the company, but I took it as an opportunity to learn: I set up conversations with everyone from the most junior junior to the CEO, to understand which roles exist, what their challenges are, what they learned, how they even got to do what they do. I came super prepared to those conversations, and for three months I checked spelling mistakes, but I also came out of there with a super, super clear understanding of which roles exist. And I remember the conversations with the product managers where my eyes just lit up, people who touch both design and the technical side and drive people to get things done, and I was just so impressed by the people who were product managers at the company. I understood product was the direction. I realized one had studied law, another was actually a psychologist, and a third hadn't done a degree at all. I said, I'll get there however I get there. Fast forward from there, I enrolled in a psychology degree, in parallel I started working at a cyber company as bizdev, and at the end of my first year I got accepted to the Wix student program. I did the student program and after that I joined as product in Wix's core product, in the Editor.

Getting into product without a technical background

Eyal David: [02:31] So let's leave the student program for the end, tell us about it then because it's interesting. Great. Tell me, A, bizdev is intriguing, B, how come they threw you specifically into the Editor, that's Wix's most painful product

Zeevi Arnovitz: [02:42] Yes, great question. I also think it says a lot about Wix's character and why Wix is an amazing company to start your career at. So bizdev: I have native-level English from home, I grew up in an American family, so that was basically the main requirement. And as for why the Editor: at Wix there's no seniority thing where whoever's senior works on this product and whoever's junior doesn't. They just find a fit and they really let you fly. If you show fire in your eyes, show capabilities, show a fit and a good fit for product, they really give you opportunities. I could say a lot more about that, but basically that's where I was placed. I was two years in the Editor, then I moved to Wix Partners, also a super cool team that handles our offering for freelancers and agencies, and recently I moved to an AI team that's doing something super cool that I hope we'll ship soon and be able to show the world. So mysterious.

Eyal David: [03:40] Okay, we'll definitely talk about AI, the most intriguing and interesting thing, which we can talk about with you from this angle and also where it catches Wix. But I have to ask you before that, what was it like being a junior at Wix? You got to Wix and they put you in the Editor with all kinds of senior folks, I imagine exactly what that was like

Zeevi Arnovitz: [04:11] So first of all, the student program is built so that you come in gradually. They also throw you into the water, but let's say, I like to think of it as being thrown into the middle of the sea but with a lifeguard right next to you. The program is very demanding. In the end they accept eight students with a lot of motivation and a lot of potential, and they really take you through the product process from zero. After that you join a team and you're paired with a senior PM who can work with you through the process. And again, that balance between throwing you into the water and being a lifeguard: they let you work as a product manager, they give you full ownership to research the products, talk to the users, make the decisions, and when you maybe start to drown or something, you really have the people right there next to you.

Eyal David: [04:40] Beautiful. I imagine it, that you came in, like you say, and they sheltered you and so on. And maybe now it's interesting for a second, when we talk about AI, how maybe the roles flip a bit, that's really interesting, but okay, we'll touch on that. In short, tell us for a second about AI and Wix. I see a lot of risks for Wix in the AI era, but tell us how the company sees it from the inside. Sure, as much as you can of course. Yes, first of all, the speaker is not the CEO and not the company's spokesperson.

Zeevi Arnovitz: [05:10] But I'll say how I see things. True, AI is reshuffling the deck. There are lots of products now where you can spin up a great landing page in a very short time. Complex websites are a bit harder, but really, Base44, Lovable, Bolt, really amazing products that I've also gotten to use. I think Wix has a few rights to win in this space. First of all, domain knowledge: Wix has been in website building since 2006, understanding how to make the very, very complex building experience accessible to DIY users. So really this combination of building with AI is something we're very, very focused on right now, and let's say we're not ignoring this threat; the whole company is going all-in and really working toward it. The second thing that, in my view, puts Wix in an excellent position is our back office. Wix isn't just a website-building platform; Wix is also an end-to-end platform for running the business. I think everyone looks at Wix as drag and drop, and that's fun and really amazing, it changed the industry completely and no competitor does it like ours, but we also have a whole back office offering end-to-end tools for running the business, whether it's e-commerce, bookings, marketing, and I think combining that with AI is a superpower no competitor has. You know, the only one, maybe Base44 by Maor Shlomo did it excellently in my view, it brings in a great back office with analytics, users and a database out of the box, but again, between that and Wix's back office there's a huge gap. And the third thing is Wix's ability to move fast, plus all the experience we have in how to build a team that knows how to take an idea from 0 to 1, which is something Wix has been doing for years. And now we really see that Wix's core product, the editors, also use AI heavily, but Wix also offers additional offerings, and lots of teams like the one I'm in now are being formed and we're building sort of startups inside the company.

The AI threat to Wix, answered from inside the company

Eyal David: [07:24] Wow, you came prepared, I didn't expect that. I actually think, I think it's a big risk, it's interesting, I won't push you into a corner. I do think the business-management side, and the fact that you can maybe suddenly bring in agents that work together with the business owner, is a very interesting aspect. Shopify is probably in the same boat too. Very interesting, let's leave that. I'm really curious, what was it like that you suddenly became, we didn't mention it, but you actually became one of the most prominent voices in the company around AI.

Zeevi Arnovitz: [07:33] Yes, so again, that's another one of the advantages at Wix. Senior, junior, those are things you don't hear much at Wix, you have PMs. And as a very, very curious person with a lot of free time, my wife is also studying visual communication and is very busy day to day, and we don't have kids yet, I was always an early adopter. If I see some new product, I immediately download it to my phone, give it my email, I don't care about anything, just to try every new thing that comes out. And already in... first of all, when ChatGPT came out I downloaded it on day one and brought it to everyone at the company, and from the start I began looking at how I bring it into the products. And the whole thing especially exploded when the vibe coding movement started. I remember I was abroad and saw some YouTube video of someone creating an app end-to-end, and for me, as someone with no technical background, it was suddenly this crazy moment of watching a person create code out of nothing with words. And as soon as I got home from that trip abroad I opened my computer and signed up for Bolt, and since then I've been really, really trying to understand how we take generative AI in the LLM worlds, whether it's Claude, ChatGPT or Gemini, into our work processes, plus all kinds of tools in the style of Board, Sabibol, Perplexity, no matter which AI tool, and bring it into the work. And the moment the guild at Wix woke up.

How a junior became a central figure: curiosity, freedom and free time

Eyal David: [09:14] When did that happen?

Zeevi Arnovitz: [09:16] Not late, I'd say. Yes, before the whole Shopify memo and the Fiverr memo, the guild was already going all-in on how we bring this in. Very, very early. And from the start, what the guild did, which in my view is amazing, is it really located the product managers for whom this is their passion, who do this at home, again regardless. How did they do it? On Slack, you know, they wrote on Slack, anyone interested, really whoever's interested, whoever plays with stuff, come. Keren and Ella and Gili in the guild reached out to me directly, and I sat with them, and it didn't matter what I'd done before or after, they just really gave me the time, immediately saw I had a lot of experience, and decided to give me a stage too. It started inside the guild, and now I'm already running sessions and workshops across the whole company, including in the UX guild, which is super exciting. And again, that's Wix's DNA: if you have something to give, they'll let you fly, let you flourish, and give you the tools to really pass this knowledge on. So that's basically it, I can also talk about that

Eyal David: [10:14] Great. I always recommend listening to our podcast at 1.2x at least, and you speak at 1.2x, so that's great. People are running to 0.8. Okay, so you told us about starting early around this, what things did you build when you started? What did you show them that made them go wow, you know, Ella and all the people you mentioned? Yes, so again, I think it splits in two. There are things I do at home

Zeevi Arnovitz: [10:46] which is all the vibe coding worlds and side projects, where mostly it was important for me to see that I'm able to build something end-to-end, including a database, including payments. It started from all kinds of random ideas I had. Again, that's also a personal recommendation I have for anyone starting to experiment with this: don't wait years until you have a perfect business model and some idea that solves some niche in the market. Think of some idea, it can be a to-do list, something personal of yours, and just really build. Build something, that's one of the things

Eyal David: [11:10] That's right, I'm with you, yes, wow, my brain explodes from this because I keep thinking about how easy it is now to build things. Me too, like, we were just talking about this with a mutual friend, Alshandri, how easy it is to build things, but actually, okay, for the first time, fine, but afterwards understanding for a second what the business model is, who you're targeting, how this thing is going to progress, who will consume it, that's so critical. That's maybe the most critical thing right now, in the next two years maybe, that's worth it, definitely. I think that now, if, as

Side projects: why they matter and why not to wait for a perfect business model

Zeevi Arnovitz: [11:46] it gets easier, so too the number of lots of niche products that will launch and flood the market. And I also think there's some kind of picture, I don't know, a picture in front of everyone's face that says, come on, we'll create something and in a second I'll have a million users and lots of money. And the significant challenge, I think, which surfaces for everyone, is marketing and positioning, and how I sell what I sell and to whom, and how I hit exactly their need and how much they'll pay. Those are crazy challenges. So I'm totally with you, but I think just building something, there's no way around it, there'll be lots of mistakes, you'll do lots of things where you say, why did I build it like that, and you'll learn a lot, or you'll learn a lot, a lot, a lot about how to build right. That's why I say: start, build something, build the to-do list app, and the second you have that killer idea you'll already have the toolset, everything you need to build it more and more correctly. I started with all kinds of lame ideas, and now in my side project I applied everything I learned, the mistakes I learned the hard way.

Eyal David: [12:45] No, no, that's excellent and clear, I'm just saying there's an enthusiasm metric. I also found myself at the beginning building some agent that just didn't interest me, so I tried to finish it until, like, okay, fine, in the end I don't use it, cool. Actually when I went and built something, even, like, to streamline this podcast, I was all over it, and then I really found myself finishing it. So find a need from real life, even if it's with an agent that's even a workflow, right? Yes, maybe let's explain for a second the difference at the most basic level. Sure. To make sure all those who are running, and maybe got off the treadmill, are hearing us and learning something.

Zeevi Arnovitz: [13:14] Yes, yes, so I'll touch on maybe three types of products, or types of things I work with. One of them is really the copilot world, which is basically some entity you create that has context about the work, that has context about the team. I'm talking now really in the context of product. Totally, totally, I'll also be presenting at the company how this can be applied not only to product, but again, it's creating some entity that's like an intern helping you with thoughts and tasks, but in the end the output of this thing is text. It's built on top of Claude, on top of ChatGPT, but really, with this copilot you can go deeper too, I think it's (unclear).

How to build a truly productive copilot

Eyal David: [14:10] So first of all, let's explain here for a second at the dumbest level the difference between a workflow and an agent. An agent is simply one where we don't know how many times it will run. A workflow is a set of actions, and an agent simply turns, basically, imagine you have some content creation, for the sake of content creation, there's another flow of a reviewer, someone who goes, another part that goes and critiques this content, say, and knows to send it back with feedback on what's not good. So we don't know how many times it will run, and that makes it an agent. So basically, the copilot, let's dive into it, what did you build it with?

Zeevi Arnovitz: [14:39] Yes, great. So I've already tried building it in several tools. At Wix we build it in Gemini. I personally use ChatGPT, I really, really like ChatGPT, but again, every two weeks there's a new model; now Claude's is really changing the game. I'll say right now that this isn't my idea. I learned it from someone called Tal Raviv. By the way, I really, really recommend everyone listening to this podcast follow him on Twitter and listen to all his podcasts; I think he's one of the most influential voices right now in the worlds of how to practically use AI as a PM. And basically, the three important pillars of the copilot are: general knowledge, okay, which, I look at it as if you just hired someone new to your team, what's all the knowledge you'd give them in onboarding so they're connected. The second thing is personal instructions, which is basically all the problems we hear about AI: it doesn't challenge me, it just goes along with all my ideas, all those things. So, so, so true, those are problems, but they're problems that are solvable through personal instructions. So we set up a project, okay, and we have the knowledge, which is everything I said we bring it in onboarding; we have custom instructions, which is how I want it to answer me: don't be a people pleaser, give me pushback, anything like that. And then under that copilot we create chats, and each chat is a project I'm working on. Okay, I've started discovery on some new project I want to work on as a PM. I really treat this thing like an intern working with me, and like a real intern, even if this intern is the most genius in the world, if you don't give enough context about what we're working on and what problem we're coming to solve, its outputs will be generic, even if this intern is super, super genius. That's why it's really important to invest in the context stage. And then this copilot accompanies me in every project I do, from the discovery level, through the wireframes and the challenges and how we plan the test, and finally how the product ships. And that's most of the meat of the sessions I'm running at Wix right now.

Eyal David: [16:43] Wow, we'll really dive in later because the work you're doing at Wix really interests me, but I want to park here for a second, and I'm wondering, I thought about this before too: tell me how AI really changed, all kinds of examples basically of working with a copilot. How does it change how you think about product, how you work on product?

Zeevi Arnovitz: [17:06] Interesting. First of all, it changes how I think in general, yes, how I look at the world. AI, I think, I'll talk for a second about something that's developing at Wix. We're starting to understand that there's a sort of stairway to heaven, a few steps through which you reach AI proficiency. So really the first step is some kind of spark, and that's something you really see in people's eyes. It's some moment when they suddenly become believers. I remember when I came to show my dad ChatGPT for the first time, and suddenly we told it, I don't know, my dad is a researcher in Jewish studies, and we gave it some super esoteric prompt: tell me how the Bible connects to the mythology of ancient Egypt. And it gave such an accurate answer, and I remember that for my dad, really, that's the spark, his eyes literally lit up, and since then my dad has been a Claude power user. So really that's the first stage, and you mustn't skip it, because that's where you become a believer. The second stage is really developing an AI mindset, which I think is what you're talking about: how AI gets added to my toolset through which I look at the world, how to solve problems.

The 4-step model for AI proficiency at Wix

Eyal David: [18:20] I was just nodding, but yes, exactly right. Totally. Take me there, so where does it start? Hold on, continue from there, but out of that, basically, how much did it change your work process?

Zeevi Arnovitz: [18:34] It changed and it didn't change. Why? Because the product lifecycle stayed the same. We do discovery, we work together with the designers and developers to understand what the set of solutions is with which we can solve the problem, after we've understood it fully. By the way, it doesn't replace discovery, it doesn't replace talking to users face to face. That's something that's very, very dangerous and that's very important for us to convey at Wix. You can use it both for scheduling the conversation and maybe for summarizing the conversation and pulling out takeaways, but basically, if at your company you notice you're stopping talking to users, that's the red flag of a lifetime. That's something that's very important for me to convey. So of course in the discovery stage, then in prototypes it's coming in very strongly now, and then in the building, of course on the dev side, it's coming in very, very strongly. But again, the stages, discovery, execution, learnings from the A/B test, those stay really, really the same as they used to be; AI simply accelerates and strengthens each of those stages.

Eyal David: [19:24] So yes, for me too, what changed is mainly constantly asking the question: okay, what can be done with AI, or where can AI fit, also in the product itself, like when I'm planning it or coming to address some problem, but also when I come to collect data or any action. Totally. Cool. So okay, go on, so you basically described the spark moment, and after that you can continue on your side.

Zeevi Arnovitz: [19:54] So we really have the spark, then mindset, which you really see in people who suddenly start looking at the world differently, at their problems differently. The third step is AI-powered, which is hard to define, but basically in my view it's a person who runs daily touchpoints, lots of daily touchpoints with AI tools, and gets output from them that's super valuable. It's not just, I don't know, give me a recipe for salad dressing when I'm cooking with my wife in the evening, but really outputs you can use that strengthen your performance and speed up the product-building process. And the fourth stage is really being an agent manager. That's already a level beyond just workflows, bringing it into daily work, but really creating agents and moving from a generative place to a really managerial place, right, we manage it, the AI performs actions, makes decisions. Full disclosure, I feel that with agents I still have a lot, a lot to learn, I'm not there yet, but I'm running a lot at our company around how to become AI-powered PMs. I want to pause on that because I also think there, what are the significant challenges you have right now with agents? In working with agents, you, right? Great question. So there are a few. First of all, I think at work, yes, in the end, at home, connecting my side project's database via MCP to Cursor, no big deal, what do I care, sure I'll do it, and it'll speed up my work too. Connecting all of Wix's Jira via MCP is a risky thing. There are all kinds of ways to do it right, but again, before we, let's say that's something I wouldn't recommend every company just start experimenting with, connecting all the tools via MCPs and giving AI agentic ability to add and delete things. So that's really one challenge. And the second challenge is really finding the use cases. Again, like, in user conversations there are all kinds of agentic things that can help a lot in scheduling the user conversations, in reminding the user before they need to join, and also afterwards in summarizing and distributing the information around the company. But it's also important to know where the places are where it mustn't replace decision-making: really talking and fully understanding the user's need. In competitor research, for example, really touching and experiencing the competing product's experience. Do Deep Research, have an agent check every day what changed on the site, but in the end if you don't go into every competitor's product and go through the experience, that's not it. So there it's really hard for me to find a place at work I want to fully hand over to an agentic workflow, and that's actually interesting: if someone has heard this podcast and has an idea, they should share it with me, very much, to look for where to bring this into day to day. Tell me, so really, in competitor research work, what do you use besides Deep Research? So mainly Deep Research. I also use Perplexity, also for the search capability; I think Perplexity really (unclear) models around search, and in my view it still performs better than Deep Research. For Deep Research I actually prefer Gemini's, it performs excellently, and I also use NotebookLM. Deep Research now gives me 40 pages, I don't know how many product managers have time to read 40 pages; it turns it into a podcast, which is a great tool, a really great hack. But again, if you don't combine it with really touching and seeing and taking the screenshots like we're supposed to, and experiencing the competitors' products, that's, that's a red flag. I'll also add that even in finding which competitors interest me, I remember we had some challenge with something very niche that we needed in the product, I won't go into exactly what it was. But from talking to my copilot, it recommended I look at some product I would never, ever, ever, ever have found, and again, that really opens your mind to how to approach competitor research. Tell me, just, I think for me yes, but I'll ask you, are there places where instead of solving, it created a new problem? Yes, I don't know if problem is the word, it's a bit of a harsh word, but it adds challenges and it removes challenges, right. Like, for example, we made a product in the Editor that lets users upload logos and get palettes for their site from them, which is something very, very hard for users. I built this product, and when I came to build it we thought of building it with an algorithm that takes the colors and knows how to add complementary colors, and that's very deterministic, right, you know exactly what will come out of that algorithm every time because you built it. And I started playing with ChatGPT, sending it the hexes of logos and suddenly seeing what it's capable of. And true, you lose the determinism of knowing exactly what goes in and what comes out; on the other hand you gain a whole world of options and of regenerate, regenerate, regenerate. So I think what you sacrifice here is really that hundred percent, knowing a hundred percent of the time what the algorithm will give me. On the other hand you cover many more use cases, you get much more creativity in the results. So again, a challenge that came up here is non-determinism: sometimes you'll get lame palettes, and sometimes the text you generate for the user won't be a hundred percent. On the other hand, the use cases you can cover are 100x what you could before. And then, question, you launched it, right? Yes, we launched it, to understand. It's something that succeeded. Yes, it succeeded a lot, we also see that the results, I'm really happy with that decision we made. That was also almost two years ago, when it wasn't yet really common to make AI products that aren't just text writing. And yes, the results in the end were amazing. No way around it, selling it, especially back then, was very, very hard, right, because one out of ten was an unrelated palette, and suddenly, you know, it's very stressful to see something like that inside a product Wix takes responsibility for. But the value you create for users by giving them endless options is so much more impactful that in the end we went with it and I'm really happy. And now we really have more than 200 AI products, AI-based, inside that product

Competitor research with AI: advantages, limits and techniques

Eyal David: [26:13] That, that, that's exactly the point I'm driving at. The problem I find, at least with product managers I work with, is the difference between, like, a cool use and a use with impact, because suddenly you can do lots of things, right? Who said these colors will work, like, right? How, how, how did you approach it from a hypothesis, and what was that like?

Zeevi Arnovitz: [26:29] So again, the product has to start, like, I think there's a tendency today to say how do we take AI and dress a problem onto it, right? Come on, come on, let's find some solution, there's AI, let's put a chatbot here, and wow, like every product now has a chatbot, which is simply crazy and isn't right. We need to look at it as good product managers: first of all understand the user's problem, understand fully exactly what the expected output is, what matters, what we're solving here, and only then start thinking whether AI is the right tool to solve it or not. And I think something we're seeing in the industry now is this reversed work: technology first and then a problem, instead of problem and then understanding technologically how we solve it. So I think maybe that's what you're driving at, and I'm with you there. So yes, it's very, very important not to skip the stage of understanding the problem and understanding exactly what and how we solve for our users. Tell me, so we'll dive back into Wix, I remember, I also know how to close these brackets if all is well. I do want to know now, someone who's going to a job interview as a PM. So first of all, Ben Erez, who was a PM at Meta, I'm also pretty sure he's Israeli, has lots of sessions on this, I really recommend watching. He's very oriented toward, like, (unclear) and how you get into Meta now and so on, but basically he really explains how to build copilots for interviews. So I'll say one sentence on it, but I recommend you check him out afterwards. But again. Same thing: create a copilot, right? What makes up the copilot is the custom instructions, the knowledge and the chats. Here I'd say, in the custom instructions, have it be your coach for how you approach the process of landing a product role. Then in the knowledge, I'd upload all the best content on the internet, and there's lots of content on the internet, on YouTube, on blogs, all kinds of things like that, and there it's important that you come and separate what's really quality content from content that's not so high quality. Why? Because today what it does is a search and it chooses which content it brings in, but your goal is to understand for a second what's really quality and include it. And then you have a full copilot you can do mock interviews with if needed, and that also advises you how to sharpen your CV. Again, the analog here is bringing it the right context.

Why start from the problem, not the technology

Eyal David: [28:55] I'd go even a step further and think, after all, we have all kinds of aspects. One, like you say, content with context that's very right for the company you're interviewing at, but there's also the emotional aspect, that's what I was driving at. It's an exhausting process, right, like, actually in that aspect I think a copilot can really accompany you beyond the data and the questions themselves. Analyzing podcasts is also a great information source, right, totally. And how would you build it, basically, would you go and use Ben Erez's or something else?

Using a copilot to prepare for job interviews

Zeevi Arnovitz: [29:28] So no, personally, for things like building some copilot, I really think the project products in both Claude and ChatGPT are very, very sufficient, right, because I create a project that has custom instructions and knowledge, and under that project I have chats. If you're not expecting some action to be performed here or some very, very complex workflow, I really, really recommend just using a project. That's how I'd personally build it, and I think that's how Ben Erez (unclear). Yes, and totally, what you said about emotional support, I think that the moment there's, also in the copilot I built for work, I also put in all kinds of things I'm trying to improve, and all kinds of anecdotes about people I work with. For example, really, for example, yes, yes, yes, totally. So for example, one of our stakeholders, it's very, very important to him that you've thought about all the tiniest corners of the product. Now, the moment that exists in the project, in the project knowledge, and the AI copilot knows it, sometimes I tell it, okay, what are the next steps we need to take on the project, and it'll tell me, listen, he might ask you about A, B, C, you should check that for a moment, even if you don't have a full answer, worth checking it before you present to him. Those are really the things where, the first time I put that into my copilot and got the outputs, they saved me from hard conversations. Not that you should be afraid of hard conversations as a PM, but that's really, really crazy value you can get from the copilot. A great way also to handle hard conversations, to prepare for them. Yes, that's another one from Tal Raviv. I'm not yet, I'm not there yet, I don't feel I'm really bringing it into my day to day, but Tal Raviv really simulates conversations with personas. He says, you'll be our CTO, I'm coming to present some project to you, and before his real product review he'll do some demo talk with the copilot. That's also a super cool tactic, and if someone listening to the podcast tries it, I'd love to hear how it went.

Eyal David: [31:38] Really, yes, I love that you're so collaborative, that's great in my view. Okay, let's go back to what you do at Wix. So you're basically running workshops, right, workshop, sorry, what else do you do basically, beyond, you talked basically about the maturity, I'll call it for a second the maturity level between PM and AI, right? So you talked about three levels, let me recap for a second: there was the spark, there was the AI mindset, and basically after that you went to the AI-powered PM, right, that's the thing

Zeevi Arnovitz: [31:55] Totally, totally, that was powered PM, which is really working with it together with the copilot day to day, and with more tools like we said, like v0, like Bolt and so on. And after that there's really the agent manager, which is the very, very top stage of AI proficiency in my view. Which is where you're now basically bringing training on what it looks like, because you don't have the, as I understand it, you don't have it in-house, right? We don't, we don't have tools in-house. Wix really believes in trying all the tools, and let's say between the time you request approval to use a certain tool until you get it, it's very, very fast. Wix really moves amazingly to really teach how to reach proficiency in this. So do you bring in people from outside? Itai Bone came and gave us an amazing session, Tal Raviv is also in contact with the company. Wix really gives a lot, a lot of empowerment to people inside the company, but also believes there's a lot, a lot to learn from outside. There are also all kinds of workshops we do at other companies. So there's a lot of focus on this at the company right now.

Eyal David: [33:08] Beautiful. Tell me, you're also building something on the side in parallel to all this, I really didn't want to forget that. Yes, yes, yes. Tell us.

Zeevi Arnovitz: [33:15] Cool. So I'm really building a platform for students that lets them streamline the process of studying for exams, lets them streamline their learning, and generate exams. Personalized. And that also wasn't the first idea. I built a few ideas, I started gravitating toward edtech; I was also a student not long ago, I know lots of friends who are students, so that's great beta users. And really, I created for myself some very, very cool workflow, and I'd also love to tell about it so that... (unclear) come on. Yes, yes.

Eyal David: [33:43] Really take us from the moment it came up and, like, you started building the first milestone, what does that look like? Actually, I'll ask you before that, how do you approach it? Do you also start sketching out the different entities? What does it look like?

Zeevi Arnovitz: [34:02] Yes. So I think something that's very important and also very, very hard, there's some... because of this effect where I write a prompt and within a second I have a product in front of my eyes that I can start to feel, there's some temptation to just run and start building, right?

Zeevi's venture: a learning platform for students

Eyal David: [34:11] Things also get lost along the way. I'm constantly jotting notes on the side.

Zeevi Arnovitz: [34:24] Totally, totally, so that's really, really the challenge, yes. And I think something that's very important, really what distinguishes the amateur vibe coders who ship a very, very broken or incomplete product from the vibe coders who manage to ship an amazing product, is the planning stage. Now, the way I look at it, I saw that many of the products function as a sort of junior dev team. They don't really have the planning stage and looking at the whole project from above; they just know how to build code. And that's also how they're built, that's the system prompt, in the end, to output code. They won't give you, you'll write them something, they won't come back with a question like, hold on, how do you want to do this, A, B, C; they'll just understand and run. And that's something that's both amazing and very dangerous, and that's why it's really important to invest in the planning stage. And very quickly I understood that if I have a junior dev team, what I'm missing, as a PM who studied psychology and was a combat soldier in the army, is a technical person. So I created myself a CTO, a real CTO in ChatGPT, and now it's already a project whose goal is basically to be my thought partner for how we think about features we're going to ship and to build the planning in the best way. So what I do: I start with the CTO. I have an idea, at first it was an idea for the product, now it's ideas for new features. I put on headphones, turn on voice mode, and start rambling to it about the project like I'd talk to a CTO at a startup, start explaining what I want to achieve, what problem I want to solve. How users, what users do today, how they struggle with it. And then together we build what's called a discovery prompt, okay, which is a stage where we want to gather all the critical information needed to actually execute the feature, from the codebase, from the database. It gives the discovery prompt, which I pass to the junior dev team, which at first was Bolt and today is already Cursor. It gets all those details and returns them to the CTO. Together we build an execution plan, which I usually like to break into steps, you can think of it as sort of sprints, and each such sprint is a prompt for Cursor that has a real definition of done and some summary I want to get back from Cursor. So of course I put it into Cursor, Cursor does the step, I do the tests, I pass its update to the CTO, to the GPT project, and that way I do both the planning well and the execution, and I have sort of two entities: one is the CTO, one is the junior dev team.

Eyal David: [36:57] I thought of two things: one, I think I can contribute, and the second, I have a question. Great. On contributing: I heard this on some podcast and it stuck with me. I used voice quite a few times, at first I felt it was interrupting me, so what I started doing, really, I don't remember where I heard it, I record myself on my phone, chewing it over, walking with myself chewing over some idea without anything coherent, and then I basically upload that file to GPT, just a file I recorded on my phone, and tell it, okay, make me a coherent document out of this, like requirements, basically. Amazing. And then you can pass that to the CTO as something much more baked, because you wouldn't sit with a person in a room right now and start ping-ponging with them; you'd want to arrive with your idea more baked, and then I think the conversation, the whole outcome, would be much higher, because you'd probably come with something more baked. And the second thing I wanted to ask, basically: these sprints you mentioned, it helps you with planning, how do you connect it to a business goal? Like, why is it tied, basically, does it think with you that way, or does it think with you mainly from the angle of what will best help complete this thing in the shortest time, what does that look like?

Zeevi Arnovitz: [37:48] Yes, I think everyone has their own domain, like in a real startup. In the end, as the product person, I'm responsible here both for the problem we're coming to solve and for the strategy: how we meet the market, how we'll do the marketing for this thing, the pricing, all those things. So of course, on the business side, that doesn't fall on the GPT, and I also recommend you not let it build your business strategy and how you approach the market. On the other hand, I'm the owner of the problem, of the solution we want to understand, and it's responsible for understanding technically how we execute it.

Eyal David: [38:37] Have you thought about creating more co-founders, one that's more on the business side, basically?

Zeevi Arnovitz: [38:48] Yes, I think basically I do a lot, a lot of conversations just with voice mode in general. Like, I noticed that's where the models are improving so much. It used to be really important to give every prompt the role: you are a senior marketer with 10 years of experience and we're gonna talk about blah blah blah blah. And today the models, like Claude 4, GPT, uh, GPT o3 and Gemini 2.5, are at such a high level that I can just start a conversation and explain broadly what I'm coming to solve, and its outputs will be high level. So I do a lot of brainstorming. Right now I don't have some entity that's, I don't know, a CMO or something like that, but that could be super interesting.

Eyal David: [39:28] Tell me, another question that comes up for me: how do you basically know how to navigate between the different products? How do you know now you're going to GPT, by which task, how does it work?

Zeevi Arnovitz: [39:38] Yes, basically, if it's a task that's small at the FE level, the front-end level, Cursor can usually handle it, and with the latest upgrade to Sonnet 4, Claude's, its performance is really at a super, super high level on the front end. The moment I understand there's going to be some back-end-y change here, the moment I understand I'm maybe looking at the structure anew, adding some new experience into the app, I'll always start with planning with the CTO, and then in the end the execution always happens in Cursor. Yes, the CTO doesn't really have the ability to write code, so Cursor is always involved. If it's something small I go straight to Cursor. If it's something bigger I start the planning with the CTO.

Eyal David: [40:15] We're getting close to the end of the episode. This was really interesting. It's amazing to me to see, really, I call you junior, forget it, that's not right, but basically you bridge this so well that it's amazing, amazing, the era we're in, amazing. What's the best advice you can give now to other people who want to get into product?

Using the different GPTs right for everyday tasks

Zeevi Arnovitz: [40:40] I have a good friend who was also in the student program, Eilon, also a very good friend and also a crazy sharp PM, and both of us had lots of people reaching out trying to get into product management, and we'd advise them. And Eilon has a tip he'd always say: do something. Yes, start, do something, build something, some business idea, all kinds of things like that. And I think today that changes to build something. I really, really recommend to everyone, if you want to be product managers: right, books are good, courses are great, build something with your hands. It's like a portfolio, right, it's so, totally. Right now I'm, like, interviewing for the student program, and there's nothing more impressive than a person who comes and shows me something they built that met users. It doesn't have to be some billion-dollar startup, but really building something in the end, I think that's such a differentiator today. That's tip number 1. Tip number 2, that's great, tip number 2, I think it's simply to stay up to date. I think that's very, very hard; on the other hand, for a curious person there are all kinds of sources I recommend. YouTube is simply a gold mine. I've been a YouTube power user for many years; really a lot of the things, maybe the smartest things I know, come from YouTube, less from academia. And also Twitter now is really a gold mine of information. I recommend actually taking advantage of being a junior, of having lots of motivation, of being super, super curious, really use that to learn. After the show maybe I'll send you, so we can prepare the show notes, a few people I recommend following.

The strongest tip for anyone who wants to get into product

Eyal David: [42:23] Amazing. Besides that, you also really recommend the Wix program, that's what I understood between the lines. Totally, totally. How hard is it to get in?

Zeevi Arnovitz: [42:35] So I won't lie, it's very, very hard. We get lots, lots of applications; in the end they accept six to eight. On the other hand, there's no program like it. There's no program like it, not in Israel, and I'm not sure in the world either. Yes, they give, they give you a lot as a student. And also something I think is very special at Wix: Wix has the ability to recruit people from all kinds of fields. As far as I know from companies like Microsoft, it's very important to them that you studied something technical, and many of the PMs there fall into some typecast of 8200. Obviously not all. At our place you can see a really wide range, both of degrees, and a lot, a lot of the men accepted to the program are combat soldiers. Like, in the war you really saw that the whole student program stopped because all the men were in Gaza. So it's really special, also for me personally as a combat soldier who did lots of reserve days, it's really important to me to help people like that. So if people want to reach out to you and so on, where do they find you? LinkedIn, add me. A lot, a lot, a lot of people helped me in my hiring process, PMs who made time and did Zooms with me and did phone calls with me, and I'm really trying to pass that on as much as I can. Just reach out to me on LinkedIn, to talk on WhatsApp. If there's something specific I can help with, that's better than, like, a general consultation. Especially reservists, I'd really, really be happy to help however I can.

Eyal David: [44:01] Wow, what an honor, you're awesome. So we'll definitely leave a link, and you gave lots of recommendations here, you'll also give them in the link in that post, but do you have some recommendation? It can be a recommendation about AI content, it can be anything, content, it can be anything that can move people toward product, toward entrepreneurship, something that, you know, three months from now you'll be able to refer to and say, wow, that's the best thing I could have said. Do you have something like that?

Zeevi Arnovitz: [44:29] Yes, I think the most important thing is to always be in learning mode. I think something that's a mistake many juniors fall into is that the moment they get accepted, they want to show everyone that they're beasts, that they're crazy and super impressive PMs. I remember I joined the role in the Editor and my peers were 30, 40 years old, really, really amazing product managers, and right away I said, I'll show them how creative I am and how good I am. And within a week I understood two things. First of all, I have no chance, I have no chance. And the second thing is that they don't expect me to have a chance, like, nobody expects me from the start to be an amazing product manager. And I really switched, a real shift in mindset at that moment, and I took every PM on the team as a mentor, really a mentor for a specific area. I had one who was really amazing at everything related to data; every data problem I'd go to him. There was one who was the most methodical person I've ever known in my life; every methodology problem, how I can approach solving the problem, I'd go to her. And I really recommend to everyone to always stay in learning mode. I think that's what distinguishes amazing product managers from less amazing product managers, and I think it's something that always lets you keep progressing and not make some hard stop at some point where you say, okay, I've reached the point where I'm a good product manager.

Eyal David: [45:37] Wow, I agree with you so much. It's not only for young people in the field, it's true for everything we described here, like, whoever rests on their laurels, so like, it needs to be said. I think it's also funny that you basically made them your copilot before AI, very nice work. And I'll also say that I think you have to stay curious, because precisely since it's so easy to rely on AI and so easy to give up the knowledge you bring from home and trust it, as we said, so precisely in that respect, learn your field in depth now and do push back on... wow, that's so important.

Zeevi Arnovitz: [46:27] Yes, a critical point here: you're responsible for your outputs. That's something I say all the time, also in my sessions at Wix. That's one of the biggest risks in the world, right? That we start relying on AI stuff, and suddenly you see something in a product review, in some prototype you're showing, and people say, wait, what's this, and you say, oh, the AI put that in. Two things. A, that's not it, it's you, because it's your output, and you must, must, must be accountable for everything and critique all the output there is. And the second thing is, you're ruining it for everyone else, right? Because someone who isn't a believer suddenly sees this thing and says, okay, you can't do things with AI, it just looks bad. In the end we all know this is the future and we have to get into it, and that's why you're responsible for your outputs, that's super critical. How glad I am you said that. How I agree, since we talked about AI problems, that's one of the most significant things I'm seeing right now. Totally. Again, true both for work levels and for meetings in general? Absolutely. Beautiful. Zeevi, what a pleasure, I really, really appreciate it. Totally, the opportunity, and again, I'm learning all day long too. If anyone listening to this episode says, man, he didn't talk about X or Y, or anyone, send me a message on LinkedIn, share with me the things you've done, I'd be very, very happy.

Eyal David: [47:34] Come on, beautiful, so you'll come back in a year and we'll see how far we've come. Bye for now, bye bye. Hey everyone, thanks for listening. If you found this podcast valuable, you can subscribe and 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 love for you to write to us. We'd really appreciate five stars on every platform, and for you to follow us so more listeners can discover us and find the podcast. You can also find previous episodes on every app or on the YouTube channel, we have links in the description. So until next time, come on, be effective, and bye bye.