AI Mind Talks #4: From Playground to Platform: Scaling Enterprise AI — with HiBob Head of AI Core Unit Yoni Friedman

Welcome to AI Mind Talks by Hybal, a real story of AI transformation. Winds, setbacks and lessons from inside the company. If you're leading AI change, this podcast is for you. Johnny, I'd like to have you with us. Thank you. I'll do a quick intro. So, Johnny and I actually started at Hybal around the same time, right? And we've been building AI from two different angles. I've been focusing on AI internally inside Hybal. And Johnny's been focused on AI inside the product. What we ship to customers, how we scale it, how we make it industrial grade. And now, as he steps into the role of head of AI core unit, that responsibility exploded essentially. 25, as you always say, was about laying foundations. 26 is about market scale and impact. Node demos, not experiments, real infrastructure, real agent systems, real value at scale. So today's conversation is about what happens when AI stops being a playground and starts being a platform. You only welcome. Thank you. I want to kick this off with, I would say a feeling that I believe a lot of our listeners are getting recently, you know, scrolling LinkedIn. And everyone sees Sass is dead. Everyone should be building. A predictive AI will replace half of the workforce. Which one of these is the biggest illusion? It's a good question. And I think, I mean, illusion, and maybe this is important for the entire conversation. Illusion is a temporal thing, which means time impacts it. I think that a lot of the predictions that people say are going to happen in the next 12, 18 months. I'm not saying they're not going to happen, but they're not going to happen in that timeframe, especially not for organizations, and especially not for our specific domain of HR tech, which is different. I think like everything else, Sass isn't dead. It's transforming, it's evolving. And maybe it's all serving to say that because I work at a Sass camp company. Hopefully I want to convince myself that I am keeping my job. But to be very honest, I think that I've been working with enterprises for the past 15 years. Enterprises don't change the way that people do. We all know that turning a small boat or turning an airplane carrier to different things. It's a whole different gameplay. Exactly. For me, I think the biggest illusion is time. It's how fast this is going to take and how easy it is. And maybe insane it in one sentence is, and this is the main point for me. The fact that we've built the technology to transform how businesses work or transform the world of work is not enough. Technology is the first step. Now you know we can do it. Question now is how do we do it? How do we systemize it? How do we integrate it? How do we turn this into a production line? And that to me is the biggest kind of, you know, I don't know if illusion, but blind spot that most of us have. Yeah, so basically like what you're saying that AI at scale is an operational problem essentially. Even a change management one. HR people love to talk in the sense of change management. It is a change management. At the end of the day, there are people that need to adopt, change the way they think, change the way they work, maybe change their incentives. It's as long as AI is operated to an extent by people. The slowest buffalo is the people. And they determine how fast this change is going to happen. Obviously, you know, that statement has something built into it that says a company that is now built completely on AI. Right. The mobile is a company by himself. Yeah, that's amazing. I don't know if that skills to the size of a company like high bob or a company like Salesforce or whatever. So there's always a blind spot. This is why I don't want to talk in the finitive terms. Yeah. I, everything I say has an expiration date. We'll try to release this episode as soon as possible for the recording. Please and delete it in like a year. Yeah. So, so the finger that, you know, the bottleneck finger blame is not pointed on the technical aspect or the better model. The finger is pointed on the organizational aspect and the management aspect of things. This is basically what you experience. Yeah. Super interesting. And I think this also leads us to, you know, the next topic, which is something that we discussed about a lot in the recent years. And basically what you always echo in every opportunity we have is that the revolution of AI isn't about the intelligence. Everyone is like artificial intelligence intelligence is the actual revolution. Basically it becomes much more grounded. It seems like it's about the integration. Right? Like what does that actually mean in the real world? Let's break it down a bit. Yeah. So if it, if it helps, there's an analogy I started talking about. And some people told me that other smarter people have said it before me. So it's not mine. But I'll act as if I acted. And I think I compared this to the motor vehicle revolution. Right? So take 120 years back. Before Ford, Henry Ford came out with the Model T, there were hundreds of different cars everywhere. Right? Companies were building them. But each version was different. It was hand-built. It was fragile, expensive, unpredictable. If you bought a car in the year 1900s, you didn't need to just learn how to drive it. You needed to learn how to fix it, how to handle it when it breaks down. So a car user wasn't just a driver. It was also a mechanic. For me, at that point, that was very similar to where we are today. An experimentation phase. You build a car, you see how it works. You'll learn from what you did. Well, the axles were great. The steering was horrible. Great. Next version is going to be like that. And when I think of Henry Ford, I think not on Model T. That was the first kind of assembly line car. But all the models before that, there's like 19 or 20. I'm not sure about the letters. But every one of those was an experimentation where he did a few things wrong and a few things right. And he learned and he did it again. When he came out with Model T, Model T was the first car that was built in an assembly line. It was the first time where every car, the first one, the hundredth, the thousand, came out pretty much the same. And so you didn't need to learn how to be a mechanic. All you needed to do was learn how to be a driver. I think that for me, the motor vehicle revolution didn't happen when the engine was invented or when the chassis was invented. It happened when someone was smart enough to integrate all the different technologies together into one single cohesive system that works. And it works the first time you manufacture it and it works the hundredth time that you manufacture it. And think of what we're building today, right? When we're building SaaS, we can take a very smart person. Most of my friends are engineers. Each and every one of them can go and build something unbelievable. But can they build it for the 5,500 clients that high-bub has? Some of them are 100 people. Some of them are 6000 people. Some of them are super advanced. Some of them are very traditional. Can we build something that works every time regardless of who's using it in which way? To me, that would be the revolution. And that's what, you know, I feel like the unit that I lead and the tech organization and high-bub is trying to do. We're trying to build something that is both generic as well as very, very customizable and personal, which is what everybody's expecting of AI. So this is like basically one of your main challenges right now. Like that's what I feel from you of this analogy. And it's pretty interesting because a lot of the noise you hear from all the areas always focus more on the technical side of stuff, right? And the ability to create something that you don't need that you can only drive it if we're taking your analogy and you don't need the mechanic next to you or you don't need to be a mechanic is basically the holy grail that you're now looking to achieve. Do you have any example like from recent days that it, you know, it caught you? the situation where your focus was tending to go to tech, but then you realize it's not the thing. I'm not building things for the mechanics, but I'm building it for the drivers. Or it's just everywhere in essence of your decisions. Like everything I think it's always a balance, right? I'm not saying we don't need mechanics. I'm not saying that companies shouldn't build. I have met HR managers that didn't know how to operate their email well five years ago and are now building with cursor. So it's the world crazy. Yeah, the world is crazy right now. And we can be afraid of it or we can be welcoming to it. It doesn't matter. The world doesn't care. It will keep going. I think that whenever I hear people expecting to see the future, expecting to have full control, to have the lovable of everything, build things the way that they want. At least for our domain, my answer is start from the fundamentals. There's almost like a pyramid of needs. And before you get to the point where you vibe code on the fly while riding the bus and building new capabilities to your companies tech stack before that there are a few fundamentals that you need to learn. The work that I need to do in the morning is the same work I needed to do two or three or five years ago. So how can I do that better before I start inventing new problems? The trust that I have in technology didn't change on the contrary actually because early days of open AI and chat GPT, I could tell them that the sky is black and that the sea is frozen and it would say yes, you know, so I think there are a lot of before we get to the future. Let's make sure that the present actually makes sense that how we implement AI to our present day problems makes sense. I hope that answers the question. Yeah, yeah, yeah, yeah, I think I think it actually answers like the first myth. You know what? Let's go with that theme. That was the first myth of like Sass is dead, right? Like this is basically the answer to that. Like there is still a long way from doing vibing something to having something actually deployed in a product that is being used worldwide by major companies. And I think another interesting thing and that will maybe lead us to the second myth if we can say that there is this feeling, I would say, and I'm saying it as someone who is, I would say considering himself as a builder that everyone should be a builder. Like everyone should be that mechanic in the 4-T model that knows exactly what's going on in the engine and knows how to activate this tool and that tool in it really today in some essence requires still that mechanic skills. In general, like are we confusing empowerment with chaos when we tell everyone to become a builder and creating like this foam of builder environment? Yeah, so let's say I think there's always a difference between what I do as an individual, how I stay relevant, how I grow and learn and if anybody comes out of this thinking I said don't learn how to build with AI, then that's not the right message. However, building at home, if I'm building a pickup schedule for my girls from school, that's a very different thing than building a payroll module for my company. I think that's to me the difference. I think that everybody, you open LinkedIn, you open everywhere, everybody's building something with AI and how why aren't you building with AI? My mom is asking me should I start building with AI and got to my grandma's. Exactly. The answer is the same as with the car analogy. I think that first, everybody should be great drivers. Everybody should know how to work with AI. Your company has, and it depends on the maturity level, has already adopted certain tools, know how to work with them, try to push them forward, ask tough questions, tell the people that are owning this, tell AI mind what else are you trying to do and they will help you build it. Be a great driver. That's 90% of the job that you need to do in your workplace. For some special people that are builders, they should be building. It's the same thing. The people that tried no-call building six or seven or eight years ago, they already have that advantage. Why didn't you try it then? It wasn't as great, but it exists. The fact that the technology became a little easier or dramatically easier doesn't mean that that's what you need to do. For me personally, I have a hard time building with AI. I get to the first prototype, maybe the second one, I get tired. That's just not my character. Luckily, I am surrounded by amazing builders so I can work with them. I think this is, to me, this is the point. Everybody's creating this formal, this fear you're not going to have a job. You must learn how to build with AI. Let's remember that most of the world isn't using software all of the time. I build, learn to build furniture before you learn to build with AI in 10 years from now being a good carpenter is probably going to be more worthwhile than being great with lovable. Hopefully that helps. Again, everything with the disclaimer of, I don't know, the future. This is how I see it. Tell me, let's take this narrative and let's assume that everyone still has that feeling because it exists today and everyone builds. Why is that a bad thing? Is it a bad thing to have everyone great builders and not great drivers? How can it go wrong? Have you seen it go wrong? I've seen it go wrong everywhere. First and foremost, I think, again, bottom up approach of creativity and building stuff is great. Then came the problem that you solved in high-bub. Everybody in sales knew that they wanted an SDR assistant to help them do their work better. Suddenly, within a month, you have 20 different assistants. Some of them are amazing. Some of them are horrible. You don't know which one is which, which one should I use. Does any of them actually consider the things that is an organization matter to us? Do they know the way that we communicate with our clients? Do they know the principles, the guardrails, the privacy, the security, and a bunch load of other things that matter? When we build, if you're not part of the tech organization and a company that has certain ways of working, you're building to solve the problem. But you have no idea what other problems you've created. That is true. That is interesting for you to sit now. When you build a lot of stuff, it creates also new problems with it. I think this is something that people miss a lot of times. In a certain point, I think, a decade ago, I worked at product marketing. One of the things that really resonated with me since then is this statement that says that 95 or 97% of blogs, the only person reading them, is the person that wrote them. Because you build and you put it in your closet and nobody touches it. You write and you put it in your closet. That's the thing. Should I be building things for myself great? Do I understand how many users are going to use it? What's their specific needs? What am I ruining in the process that hasn't changed? If we take it to the world of HR, when you make a mistake in a char, if you're building a marketing campaign tool and you did a mistake, you've lost money. The risk tolerance. Exactly. You lost money. It's okay. If you screwed up in HR and sorry for my language, if you made a mistake, that's a legal, that's a brand, that's a stock value implication, those things can be destructive. We need to think who are our builders and how do we teach them to build correctly? If we make mistakes, we're going to pay for it. That's super interesting. I think also the message I believe that you want to share, you don't want it to be interpreted as if you're focusing on how to be a better driver than a better builder than you don't have impact on the car. Let's remember that you can [BLANK_AUDIO] still be the best driver of this car and influence almost entirely the next model, right? Like this is something like, let's distinguish between the ones that are defining how the next model should look like. So a lot of times, you know, when you say about yourself, I'm not a builder, you know, I'm always like, wait, but when he says builder, he's talking on the, it's not that you're building every day, literally, but not hands on. And I think this is like a key difference that it's important to understand, like you could still have huge impact influencing the how it's gonna be built, rather than focusing on I will be building it personally. - Yeah, I love this. This goes down to the point of a one man company, right? When we're thinking of builders, everybody selling the idea of a one man company, you know, from my experience in life in general, when you're alone, you don't necessarily get positive or negative feedback and there's a lot of different problems. My role has always been, I think I'm a good user. I think I'm very critical of what I do. Think of it in cars, right? You let Lewis Hamilton drive the worst car in the world. You're gonna get great feedback. You let me drive the best car in the world. You're probably gonna get an accident. But this is the point, right? I don't think that the way that we've built software and we've built tools in the past needs to change. I think that the privileged few that can build completely by themselves are great and hopefully, you know, my daughters will become one of those. But for most of us, I think the way that we build software the way that we build solutions is still the same. We do it as a village. - Interesting. - Okay, that was a lot to grasp. So everyone should build, the answer is no but. And the but was exactly what we just discussed. I want one second to jump to like a straightforward shooting question without much context, but I think it will lead us to a lot of context, something that we hear a lot and it comes together with Sass's dead a bit, which is dashboards. Dashboards, interfaces, something that is so associated with Sass in five years. And I know you always like you have like this distance lens that you don't wanna look as an oracle too long. But in five years, will I still log into dashboards, you believe, or no? - So I think the answer is, for me, I mean, you know the answer. I think that, I'll try to answer this differently. I think that not everything should be done with chat. I think that the visual aspect of whether it's dashboards it could even be a spreadsheet. Some of these interfaces, they might feel outdated today when we can start talking to our phones and soon talking to a necklace or something like that and doing everything with it. Chat or conversational interfaces don't always equal productivity or efficiency. Sometimes they're more friction than they are and which is why to your question. I think we'll have dashboards. I think we'll have customized interfaces that are very visual. I think that this thought of everything chat GPT or Gemini is gonna absorb everything. And we hear it from clients, right? They wanna know that they can communicate via Slack with our system. And the answer is, again, yes, but do I want them to be able to talk to my system from every other interface? Yes, definitely do. I think that just the conversational interfaces enough? No. - But they do understand it. - The customers, I think, or they're like into the hype and like, we want everything through chat and you usually feel the one that is like popping their balloon of like, hey, are you really, like you think you can manage payroll with chat? Like do you really wanna take that risk on? - Exactly. So this is the point. I think I enjoy being the Mythbuster. (laughing) And I think with very simple examples, we can clarify this, right? So if I need to approve a day off for a team member and let's imagine a month from now, my phone just pings me and says, hey, this employee wants to take a day off tomorrow. Is that okay? My answer is yes, I can also ask a clarifying question like, is there a reason not to or do they have something to do tomorrow? That makes sense. What happens when I need to approve 30 days off? Right? Do I start going one by one with conversation asking the same questions? - Interesting. - Or like that becomes friction. As opposed to that, the way that we've done it, five, 10, 15 years ago is give you one simple screen with the list of all these people and you have, you can accept or not accept, you can do bulk approve, you can apply filters to see who might require more attention. Those are simple things, right? Not all the interfaces and I'm gonna take it to the extreme level, right? Let's take it to the car, right? Would you want a conversational interface with your vehicle, right? If it's a self-driving car, then sure, I wanna tell you where to go, but if it's like a car I'm driving, and I need to say left, right? Not that much, right? Break, that doesn't work, right? Some interfaces shouldn't be conversational by definition because they add more friction that they add value. So to that point, yes, we'll have dashboards, yes, we'll have forms and spreadsheets and things of that sort, will they be as much? I don't think so. And I'll tell you more over, for me, the our goal is to have these dynamically built based on context and intent instead of the way that they are today. So today I'll give you a form that I've built based on research of hundreds of users, and that is the form, right? And you might be able to configure a customizer to a certain extent, but that's it. I hope then in two, maybe even a year from now, that interface will self-generate based on the context, the intent, and maybe what the hundred users before you did so that we can learn how to do it exactly. So it will, and it will go hand in hand with conversational, but you need a visual. There's so many things that require a visual or require touch or anything of that sort. That's interesting. So you're describing this future of a hybrid approach. Like it's not one or zero, like it's not chats versus dashboards. It's gonna end one way or the other. We're talking on basically the ability of the user to chat, I would say, or to interface and at the right moments to switch between those two capabilities at the right moment. - Yep, definitely. - This is also something that when you're working on the future of AI at high-bob at the product, is that something that you're also pushing inside the product? Like this is also how we're envisioning the roadmap of high-bob looking forward, like this mixture and combination. I know Schmulik, our VP of design, is talking with you a lot about it. - Yeah, so luckily we have great, an amazing design team. They've revolutionized HR 10 years ago, and I believe we're doing it today. - Crazy team, literally. - And yeah, that is the idea, and we already see it, right? So today, it's very basic, but we can generate charts and tables and things of that sort based on context. Think in a few months from now, I'll talk completely different. We are going and generative UI is not something that we've invented everybody's talking about it. Think of now there's like LLM apps and so on. - That's the MCP UI. - Sorry, MCP apps, sorry. Yeah, so we're going in that direction. Definitely, again, I think it is slightly more complicated for us HR systems are not just basic interfaces and some of them are quite complicated. Some of the decisions that you make in our system could be dramatic, right? If you're running a compensation cycle for 1,000 people, the financial impact on the company is amazing. If you're sending a paycheck for 1,000 people right now. But yes, I believe that's exactly where we're going. And I think we're not the only ones, right? I think that's where the industry is going. - Yeah, it's going to be exciting to be honest. - But also confusing, by the way, just sorry for interrupting, just one point. - If confusing for you or for the user. - For the user. - For the user. - I agree. - I am used to opening the same screen and seeing the same interface. - That's such a true thing. Also, like me as a user, I feel that I would know blind spot downloading an app or getting a SaaS platform. I knew already what to expect right now. It's like, wait, what am I going to get? I'm going to get a dashboard. I'm going to get a chat. I'm going to get a mixture. Are you going to force me to use only chat? - Yeah. So the button was here. yesterday. Where is it now? Right? Those are simple things that we need to consider. The interfaces have evolved, but I think, you know, I recently read that the problem with conversational interfaces was that people started like becoming perfectionists. Right? They got to the answer and they kept iterating. They kept talking to the AI because they wanted to get the rabbit hole. Exactly. Right. And sometimes it's just like, you got an answer. It's pretty good. Go with it. Keep going. Right? Well, that's that's so important. You know, this this tension between you could easily fall to the AI slope. Let's say when it comes to like over over using AI. Yeah. I'm using it. And in a sense, one one last point of it that one of the problems is that the people that are building systems right now are savvy, advanced, creative users, but they're not building for users like themselves. They're building for me. They're building for the, you know, the non-technical people that are trying to get their job done efficiently. And there's really a conflict or not a conflict, but like a gap between what they expect users to be able to do. And what users actually will be able or will enjoy doing. And we need to keep that in mind. There's a certain level of humility that we need to maintain when we're building. We I use that matter for before. We're building for the slowest fast buffalo in the herd. Right? We want to make sure that everybody can adopt it. Even the poor drivers should still be able to drive the car. I feel like it's also like there is this paradox of what you just said that we can also add a third layer to it. And that's something that I believe you find yourself a lot dealing with is that the slowest one in the pack, usually also ask for AI without knowing what it means. Like they're also part of this fomo. They're also part of like customers are coming and saying, I want everything AI. And you know, usually there is this no saying of like the customer is always right. I feel like we're living in times where you are becoming this type of a guardrail of the trust of the users. Like they don't understand a lot of times that trust is more important than a 10 aha moments in a day of like, oh, the AI did me that and it solved me that. But this one time where trust breaks, is it still worth it or not? Yeah. Yeah. So I agree completely. I think that we, this could be clients, this could be leadership, this could be the market. You know, everybody saying my kid built this amazing thing with loveable last week. Why can't we build that into the product? Well, because your kid doesn't care if it fails or if it doesn't fail. Your kid isn't managing millions of dollars in payroll or in or keeping, you know, PIIs of thousands of people that if you know, if it leaks, you're going to get sued. I think that we keep thinking if I take this to the analogy of the vehicle, everybody's thinking about the engine, the technology, how fast can it go? How, you know, how strong can it be? If you give me a car and you say, I've doubled the size of the engine, it is twice as strong. The first thing I'm going to ask is, are the brakes as strong as well? Right? Is the steering well going to handle it? Will the axles break when I hit the certain speed? And when we're in hype, are those questions are being asked at all? Like no one cares about them except of us. Like as a company who's building this service that customer's trust in. So I think luckily, every company has a lot of smart, reasonable people, not that, not that the other, people that are expecting progress aren't smart. But I think everyone has, you know, there's a CISO, there's a CIO, there's people there that are going back to, hey, why do I have this system? I have this system because I'm managing a risk, because I'm protecting assets, because I'm trying to run business processes in a predictive way. And it doesn't matter how much sparkle and how much flare I put on the system. If it doesn't do the job, if my trust level decreased from 100% or from 99% to 90%, I'm not going to use it. And I don't care how good it is. And that's true for most reasonable people. You'll give me a car right now. And you say, this is the fastest car in the world. It breaks only once every three times. I'm not getting in. Right? And some people might, right? But I think most people won't. I'm trying to grasp it and to push us forward to a different angle of that. So like this is the angle of people should be aware of the risks. And most of the companies who provide software to people should balance the, I would say, impact of AI versus the trust. But there is also like the side of like predictive AI. I think this is something that I feel a lot of times today from different products that I use that I'm getting like 10X notifications and alerts that are pretty smart. I have to say, but I didn't got those alerts before. And to be honest, it haven't been a problem or took CPU of my mind before they started to push it to me. So like, this is also something that is pretty interesting that's happening. And I know that it's something that you're thinking about a lot, right? Yeah. Yeah. I'll start by saying that and this is something I think I actually invented the sentences, the agent that cried wolf. I googled this morning, I couldn't find anything. This is mine. I'm going to trademark it. I think that everybody's expecting predictive AI, right? My day is a mess. I have tens of tasks to do. I don't remember them. I keep them all in my head. I, you know, and I would just love for an AI to just tell me, what do I need to do right now? What's coming? You know, what do I need to think of? And that might be again, great for our regular normal lives. I can tolerate getting a prediction that might not be real. When we take it to the world of businesses of enterprises and specifically of HR, I am afraid of providing predictions to my users. The first thing is again, this agent that cried wolf, will I give them predictions that are not true? Will I go to a manager and say, hey, I think that based on these and these reasons, this person is about to leave what happens if they don't leave. Did I start a conversation that they weren't even thinking about? Right? Think of the mistake, the cost of mistake. Think of the ethical aspects of reading into something that might not be there. And lastly, I think that this is, and you talked about it, this is an idea of trust, I can get 100 perfect predictions. When the first one that is wrong comes in, I'm not going to trust it. Now take, let's take it back to how people experience AI today. We're all doubtful. We know that AI elucinates. We know that AI is trying to satisfy us. We know that AI is just trying to keep us in the conversation. Stay in the conversation. I don't trust it to that extent. I ask very specific, explicit action. It's sorry, questions. And I take each and every one of them with a big grain of salt. So I think, again, I talked about the pyramid of needs. When AI is going to be consistently great with solving the problems I'm trying to solve today, I will be open to start exploring new problems that I wasn't aware of. But don't everybody's trying to get to the end. Yeah, right? Start from the end. Yeah. We're not there. We're just not there. And so I think will predictions, will AI predictions come in yes in certain places? There are certain areas where I think AI shouldn't be involved. We've learned that the best example is around talent acquisition around hiring. AI should help us when it comes to hiring, when it comes to the future of people. Am I going to hire this person or not? Are they better than someone else? I think that while AI might get to a better outcome, sometimes AI is going to consider things that are not important. Like demographic aspects of that person, where I, for myself, I'm not necessarily interested in having them as a factor. That problem existed before Gen. AI, but Gen. AI is making everything so much simpler and it's scaling. - Yeah, scaling the problem. - Yeah. - Yeah, and also, you know, like empathy and taste, especially when it, like all those areas that requires empathy and taste is something that you wouldn't want to raise a flag and say, "Hey, I can predict you what's the right course of action because still, and that's my prediction at least for the near future years. I don't see it overcoming human taste and instincts and intuition, you know, all those stuff, there are very vague, but are super valuable, especially in environments like HR. - Yeah, even in the extreme cases of a one-man company, there's still a human in the loop, right? There's still that one very talented person that has instructed their army of agents to know where to ask for approval or ask for feedback or ask for course correction. Those things still matter. I am trying to not imagine how our reality would look like when it doesn't happen, but for the near future, this is where we are at and I think that if we want people and AI to work in harmony together, the first thing we as people need to do is trust our AI. I will be happy. I keep saying this to clients and I think it resonates pretty well. I don't wanna build the smartest AI, I don't wanna build the fastest AI, I don't wanna build the coolest AI. I'm also probably not equipped to do that, right? We're not open AI or not Google. I wanna build the AI that says no when it needs to say no. If my AI knows my AI, if high-bobbs AI knows when to say you're not allowed to do this, you don't have the permissions. This is against company policy. This is against EU regulations and so on and so on. I will be so happy and I'll be proud of what we've built. - Amazing take. I'm jumping to a high-level question which will lead us to the last part. What's a one belief about enterprise AI that you think will look back at in three years on this episode, me and you with some popcorn and laugh about. - Wow, that's interesting. In three years from now, listen, I think that the one thing that is going to dramatically changes that everybody's going to be AI savvy, everybody's going to be great drivers. We're still, I am still thinking about how do I take a person that is never drove a car to help them drive a car. And I think that creates a certain complexity or sophistication level to the solutions that we're building. In three years from now, I think that definitely the type of organizations that high-bobbs work with that are modern, multinational. I don't think we're going to have a lot of slow buffaloes. I think that the slowest buffalo is going to be probably considered if that person comes into three years to now. That person is probably going to be considered a very valuable adopter. And I think that is probably the one thing that is going to change. Still, that doesn't mean that people are not going to be needed. But work is going to change the way that we work is going to be redefined partially by us, partially by our competitors, partially by the users. And if you work in a modern company in three years from now, you don't have the privilege of not using AI. That's a ticking time-up. - I want to wrap up this episode and suddenly I got this realization of-- - You want to talk to me more? - We're building. That's for sure. And we're going to talk a lot more just off camera. But we're building a high-bub, a software for people, for workforce enterprise. And this huge transformation is on people, on enterprise. And in the last 35, 40 minutes, I swapped, like I got two times, three times, four times in my head, like, oh, I would love to take Yoni's role and build. And you know, be the one in the middle of the heat. Building this transformation, helping companies go through every transformation. It doesn't matter if it's AI, Cloud, mobile at the end of the day, it's all about people. So I'm envy or I'm saying, you know what? I'm happy that Yoni is acting that and I'm acting my part. You know, I'm just every time that I talk with you, I'm amazed about the challenge that you took. And how you and your amazing team and everyone, all the village around you, and what we're doing at high-bub to help those companies. And I think our customers feel it, know it. And we share a lot of that knowledge with them. This is just a small part of it. Thank you. I want to say thank you. That was my thank you. We'll finish this episode. Like always, we had, of course, our agent listening to the entire conversation, a quick 60 seconds chat. Let me see if the questions are ready. Okay. Short answers. No explanations. First instinct. Yoni, let me put one second, 60 seconds because in the last episode, it suddenly became two minutes, stuff like that. And I got complaints from the listeners. So I'm on it now. Okay. Okay. Let's go. Most overrated AI trends right now. Creating pictures and videos. Most underrated AI capability. No taking. Biggest mistake tech leaders are making. Rushing into things. One belief you changed this year. I don't know that AI that you can, you can have a company completely be run by AI. Interesting. One thing enterprises pretend to understand about AI, but don't they think that they can break all the silos with AI and potentially they will be able, but not right now. One word that describes AI adoption today. Chaos. One role that becomes more powerful because of AI. We talked about it earlier as the integrators, the forward deployed engineers. Amazing. 60 seconds. Thank you very much, Yoni. Thank you. (upbeat music) (upbeat music)