Latest / You've Been Heard / 381- Solving Wrong Problems Perfectly w/Dima Syrotkin

Transcript

381- Solving Wrong Problems Perfectly w/Dima Syrotkin · You've Been Heard

• 0:09
  Speaker A: You know, we talked a while ago. All I remember is that I really just liked it. It was a very, very exciting talk. And I remember, let's talk about delivering on like a ERP project ahead of time. And you're like, what do you mean ahead of time? Let's just talk about on time.
• 0:24
  Speaker B: Yeah.
• 0:25
  Speaker A: And so I was like, oh, touche. So, so yes, yes. I mean, I don't really know where to start. Why don't we just start off with like. Because we wanted to talk about real AI use cases. And with all of the, I don't know, I think 12% of actual venture capital money being spent out there or invested in AI, only 12% of it's actually making money or doing anything worthwhile. And I don't think we're really replacing people and humans yet, even though we say we are and that's not happening. It's just an excuse to lay people off. But, you know, so we do need to talk about actual real AI use cases. But I don't. Give us your, your background. How'd you get started in this weird world? This weird world we lived in, other than being born into the technology world?
• 1:13
  Speaker B: Yeah. I was 19, I joined a student organization called ISAC, one of the biggest student run organizations in the world. And they have like 120,000 members worldwide. Something. And because it's so large and it's student run, they have to educate people a lot. And there's a lot of conferences and trainings and most of them are functional. They just teach you how to do the job, which is they send in students for internships abroad. That's kind of what they do. But beyond that though, there are some conferences that are more on like personal development. And I went to one of those and it was like, you know, who are you and why are you here and what are your fears and what are your dreams? And they run simulations and then they make you reflect on how you act in those simulations. Because typically the way you act in a game is also the way you act in real life and the choices you make there. And they had this like small groups of 12 where you would like share your life story and then kind of like become really intimate over the week with those people. And, and that changed my life. And I was like, wow, that, that's very powerful. And so I started, you know, and.
• 2:24
  Speaker A: I dig in on a little bit. I, I, I just, I gotta know, like, I gotta know what was like. Because many people are afraid of a lot of things and that could run their life and then people realize that they don't need to be afraid of things, or they do need to be afraid of things and then just confront their fears anyways and say screw it and do it because we're all afraid of what? Whatever. I don't get this afraid of success thing. I think you don't have success because you're afraid and that cripples you. I don't get this like you're afraid of succeeding, but I think anyone would love to succeed. But I. Maybe that's just like something I'm not getting in my head. But what, what, what was the fear thing or what was the, what was kind of like this big driving factor? I'm just curious.
• 3:05
  Speaker B: Death. Fear of death. I had some challenges with health early on when I was like 15 or something. And then I think, you know, the urgency it gave has some like, good positive sides to it. At the same time, if you have kind of thinking about it every day, it also gets a bit too much. Yeah. So that, that was on the fear side. And then. But my big discovery, even bigger than that was just that like I was like, oh, like first of all, I like this leadership thing, whatever I did in the student organization and like leading this team of like you 18 people and founding an organization from scratch. I like all that. And then I was also like. And by the way, I also suck in emotional intelligence. Like, I didn't even know this thing exists. And then people told me like, yeah, you know, like you can actually be aware of what you feel other people.
• 4:00
  Speaker A: You can be aware that you're kind of a jerk sometimes.
• 4:03
  Speaker B: Exactly. That's me. Yes.
• 4:06
  Speaker A: I'm about 50% emotionally intelligent.
• 4:10
  Speaker B: Yeah. I think I wasn't at zero, but.
• 4:12
  Speaker A: I was at 10, no go. I was thinking of what my kids would say. Yeah, let's see. When I worked at Starbucks, my emotional intelligence, like quotient probably like went high because, you know, they. And was back way back in the day, back when Starbucks was trading really high. And it was kind of like the heyday of like, you know, Starbucks. And we still walk into the store and smelled the beans and like it wasn't quite so McDonald's esque yet. Was still, you know, was still like. Yeah, it was still like an experience. It was still like an experience, I guess.
• 4:46
  Speaker B: Nice.
• 4:47
  Speaker A: And so they're very big on, on. They gave us a ton of training. Like they really. They gave me probably the equivalent of an MBA bringing me in to teach me about P L's and EBITDA and flow through profit and all this different stuff that. I don't know if they really do that for a store manager now. I think they probably just like, you're a. Meets expectation, needs improvement type of thing. Here's a book with some, you know, maybe they still do, but I think they are investing a lot in their leadership back then. So I got a lot of benefits out that Franklin Covey stuff. Courses. Remember some other. They sent us a lot of courses. It was really, really awesome.
• 5:21
  Speaker B: Yep.
• 5:22
  Speaker A: And I remember reading Emotional Intelligence, and at the time, I had maybe four kids, so I could still teach and have, you know, I was teaching my kids and let's do goal setting and all that stuff. And then I ended up with eight kids eventually. So I think my ability to divide time equally kind of like went out the door and then it became back to more of like a carrot and sticking and mediocre. I mean, meritocracy. A meritocracy of like, you know, So I. It may. I think it. I think it ebbs and flows. However, when you have kids and you have girls and boys, all with different personalities, you realize that some of them think you're a really good listener and others think you're a terrible listener. So I think trying to be rated high on the good listener scale with all of your kids would mean that you have a higher level of emotional intelligence because you've learned to connect, discover, and respond with various different personalities.
• 6:25
  Speaker B: Yep.
• 6:25
  Speaker A: Does it sound. Does that sound somewhat emotionally intelligent?
• 6:28
  Speaker B: It does. It does.
• 6:30
  Speaker A: Or working or when you travel in other cultures, because I'm. I'm assuming you've traveled, clearly you do have an accent and. Or I have the accent. There you go. There's the emotional intelligence. It depends on from what frame we're looking at this. Right?
• 6:44
  Speaker B: Yeah.
• 6:44
  Speaker A: The white American accent. You realize sometimes that other cultures, you really have to kind of like, decipher and figure out what's going on. What do they find is rude? What do they find is normal? Like, you know, you give someone the middle finger and, I don't know, somewhere else, they. They're like, cool, man. What's that like? Or people point with the middle finger and you're like, in America, that's a no. But in other cultures, you might do something with your finger that's like, really offensive. So I don't know. Sorry, I digress. But that does have a lot to do with leadership and definitely technology, where many of the technology people are kind of looked upon as drones and not emotionally intelligent and hiding in a server room or code or Hiding in a dark room, coding and don't talk to me, I'm smarter than you type of thing.
• 7:28
  Speaker B: Yeah, yeah, yeah. And the interesting thing was basically I started a bunch of businesses in this, in this field, right? So initially like a corporate training marketplace, then I worked on like measuring effectiveness of corporate training. I worked on matching startup founders to help them kind of help each other and coach each other. And I learned that like, you know, there's a very interesting problem that is a multi billion dollar problem which is looking at it not from a kind of individual perspective, but looking at it from the organizational perspective. So what happens on the organizational level? What does development, emotional intelligence, all of those things, what do they mean on the organizational level? And fundamentally, if we look at the job of the C suites, I think of their job, I think there's two, two, two jobs. Basically there's strategy creation and strategy execution. Like you need to know where to go and what, what to do and then you need to actually do it. And that whole change management, right, that you also put in the title today, that is the execution part, that is, you know, managing change, managing that, you know, implementation of that new strategy, whatever that is. Right now most of our use cases are on like AI adoption. So, so most companies are now like, hey, we need to implement AI. So that's one example. But it can also be things like, okay, we want to be more innovative and you know, bring new products to market and things like that. Or it can be M and A. We just bought some companies because we think that, you know, it will kind of help us accelerate and sort of increase. You know, one plus one will equal three, but also notoriously, most MNAs failure. And that's because of the execution part. Me.
• 9:22
  Speaker A: Okay, I gotta give a shout out to Alex Hermosi because I flew out to Vegas to see him a couple weeks ago and it was life altering. Thank you, Alex. You're not gonna listen to this. You're not gonna listen to this. You're not gonna have time. But maybe I'll throw, you know what I'll do is I'll throw this out as a Google review for Alex Sharmosi. I'll be like, we reviewed you on your podcast. Let me tell you about what Alex's team is like. Once you get there and you meet his team, you could care less about meeting Alex Hormosi. That's how good his team is. Every single member of his team is the best of the best, the absolute best of the best. And that's one of his, like hiring Strategies. It's like you either like hire like, you know, young and hungry and like really like, you know, or you just have to hire the best. We're going through a rebranding now. We're not going to be dissecting popular IT nerds anymore. It's going to be this massive reveal next year. But because of that, I hired the best. It was not cheap, but David Breyer will give you a shout out as well. Who was the guy that branded the company Know Before. For anyone that knows the security company Know before is now going to be. Is. Is our branding expert. And he's, he did a great job making sure that IT leadership is really, really felt and, and heard. And that's a little bit of a foreshadowing. But, but as far as Alex, the reason why I thought of this is when you said creation and execution because one of the things that he likes to go over, one of the things that he talked about is, you know, solving problems and kind of going like, what's your number one problem at this time? Right? So if a CEO's idea is to, you know, create this like strategy and plan and then execute on it. And part of it creating that strategy and plan might be solving a problem or solving a problem in the company, depending on if he's coming and taken over or something like that. But one of the worst things that you can do is solve the wrong problem. So meaning, like that problem might not have any trajectory towards what you're trying to, what you're trying to accomplish as a company. It might have, no, it might not have no impact on the bottom line. It may have no impact on numerous, numerous things. And then you go to solve this. Like, if you're solving the wrong problem, then moving on to execution, now you're executing on solving the wrong problem. It's like this really like kind of of like snowball effect, if that makes sense. So I would love to ask you about from the creation piece and the execution piece is like, well, okay, if we're executing, we might be executing really, really well on solving the wrong problem. Well, how do we know that we're solving and executing on the right problem to begin with?
• 11:57
  Speaker B: One of the elements, right. Is. So I think that, you know, if things go really, really well, if we look at like the current stock market, if we take out the magnificent seven, most big companies, old companies don't really grow. They're quite stagnant. And I think part of the reason could be their size, right? At some point when they get to certain Big size, they just like, they stagnate. It's like innovation stops, there's too much bureaucracy. Things don't go anywhere. And I think the solution is not that. I mean, yeah, part of the solution could be just smaller companies, but I think part could be in the future is that you would have C level and you'd have individual contributors, and the middle management in between will be AI and the coordination will be managed by AI. And so the question is like, okay, what does middle management do? I think fundamentally middle management has a couple of jobs on one hand. You know, it's aligning people around that new strategy and telling them, like, hey, guys, this is what we're doing. Do you need any support? Let me coach you, Let me help you. This is where we're moving. And what are your goals? Okay, Your goals are not aligned. They should be more aligned with this. So all of that. There's the other side of it, which is they also help C level to collect information from the front lines.
• 13:23
  Speaker A: That's exactly what I was talking. Tracking.
• 13:26
  Speaker B: Yeah, exactly, exactly. Like, what is really happening in the company and what are the best ideas that the people on the front lines have? Like, what are they saying is breaking? What are they saying? Should the strategy be. And of course, like, you know, they might be wrong because they are looking at something very like, myopically at their part of the company only. But they could also be. And they could also spot something that C level would never notice. And so, yeah, talking about strategy creation and how do we ensure that we are actually looking at the right problems? I mean, I don't think, like all of it is quantifiable, but I would just say, like one of the factors is exactly that, which is also what our product does, but like helping this new level understand, like, what's.
• 14:13
  Speaker A: What's happening on the shop floor, basically, and delivering on. Delivering on what? I'm assuming numbers and reporting and feedback. Super awesome how this kind of like, it's kind of in parallel to what we're talking about. But you mentioned being in an environment, the environment that you kind of like surround yourself with. From a. From a creative business standpoint. Whatever you were talking about, you mentioned moving and how a move really kind of like affected you. I just was hoping you could talk for a few minutes on that because I just got off a plane yesterday. I'm in Morocco, so I did kind of like maybe the opposite.
• 14:51
  Speaker B: Yeah.
• 14:52
  Speaker A: Not. Although Morocco is in the top eight African nations that's, you know, has like a. Whatever growing GDP or whatever. That is. And I might disagree with you just because I don't. I'm not like a. I'm not like a Silicon Valley guy. But you are, so. But, but I definitely, you know, if I. Maybe you can convince me.
• 15:11
  Speaker B: No, I think, like, fundamentally, just like my experience has been that first of all, like, it's. I, I like making friends, and it's easier for me to make friends in an environment where these people are a little bit more like minded. They don't have to be the same, but they are also ambitious. You know, a lot of smart people here, creative, also weird. Like they think differently. So. So this like. Yeah. And the Bay Area was just like the place where I, I found that concentration to be the highest. And I, I find that beyond just like enjoying my life more and like making more friends, I found that it also was a very learning experience being here because there's a lot of people here who made it already. And it's like the way you, like travel to see Alex Hormozi. It's like, you know, I kind of feel like I'm surrounded by Alex's formosis here, you know?
• 16:18
  Speaker A: Yeah, I get it. Although I'm not going to Vegas. I'm not moving to Vegas. I should, God willing. He's in Vegas.
• 16:25
  Speaker B: Oh, wow. That is quite a place to be. Quite a place.
• 16:30
  Speaker A: I mean, I just, I can't see myself moving my business headquarters. I mean, I'm sure there's. I'm sure he's answered this before and, you know, shame on me for not knowing, but I'm sure there's like, you know, like legal reasons and tax breaks or something like that. I'm sure there's, you know, some reason. It's like, you know, I don't know, Musk moving to, like, where did he go? Like Texas or something? Where did he. I don't know.
• 16:52
  Speaker B: But then he bought Twitter and came back, so that's.
• 16:56
  Speaker A: Yes, I digress. So how do we get on to executing and managing the cost of change inside companies that's really cost billions every year? And how could we say, take an insane ERP project that is, I don't know, through an M and A or something like that, and actually deliver it on time and on budget? That is. That's a crazy, crazy idea. That's what everyone's called. And everyone believes. Everyone starts off believing that. Everyone starts off believing we will deliver. I believe the vendor, he will deliver on time and on budget. And there might be some 5 percenters out there that do that.
• 17:45
  Speaker B: Yeah, I Think that like all the old tools should stay in the box. But I would kind of propose adding one more tool. And basically by all tools, I mean things like, yeah, consulting that helps you figure out if you should even do it. And then, you know, it will be like all the workshops and trainings and elearnings and emails, all of that is probably good and probably should stay. At least for now. Where I'm contributing is saying, okay, so how do we support the individual? Right, so having this AI coach, AI consultants that can talk to thousands of people and provide personalized support that both takes into account their needs, but also organizational strategy. And so it helps them understand what the hell the company strategy is, how it relates to them, helps them set goals, follow up on them, educates them. So for example, if we work with like AI or erp, what can AI do and what it cannot do, what should they expect? And then helps them brainstorm, etc. Even like emotional issues, like many people with AI specifically, like many people are afraid of it and afraid that they're going to be replaced and things like that. So it could help them understand like, okay, is that valid here? And how should you think about AI? Could it be an opportunity instead of being a threat? So that's level one, Level two is the group level and team level. And that's where we are not doing much yet, but we're, we're really excited to do work in the near future and that would be things like connecting people. So I ran like focus groups where I asked people in the companies and I asked them like, if, you know, you work with your AI coach on the problem and it says, hey, besides you, there's someone else who's struggling with the same problem. Do you want to connect with them and see if, you know, you could kind of figure it out together. So like a staggering 85% of people said they would press. Yes. And that wasn't the case for every feature. This was like the highest rated where people are like, yeah, I definitely would want to. And they told me they were like, look, you know, often in the big company I have no idea what other people are working on or what they're doing or like. And often we are repeating the same stuff where we often later realize someone else already did that job. And also it kind of feels lonely because even though it's a huge company, you have a lot of colleagues, but you kind of don't know most of them. And so you don't quite connect very, very frequently. And this kind of trauma bonding around the same problem could actually be one way to connect and then just.
• 20:26
  Speaker A: You're blowing my mind right now because it's kind of like has even from a sales really. So for all the CEOs that are not listening to this, the CTOs and the IT directors and managers, you could deliver something that could actually probably really drive sales success, which is how many salespeople deal with a very particular type of objection or block to a sale. And there's another sales rep in the organization that has dealt with it very, very successfully and the new guy doesn't know how to deal with it. I didn't even know that other person exists. Or I mean really from like kind of like that standpoint or like what emails? I mean there's a lot of things from sales and marketing that can get very, very, very tiring and exhausting. Split testing things, sending out different emails, all these different things that marketing people just kind of throw up against the wall and just, you know, hope things work. And I'm really hoping they can apply that to LinkedIn. So I stopped getting really, really annoying invites that are like five paragraphs long. Yeah, should be able to do that. What have you guys seen that's been very, very successful. That's driven the needle. So what AI use case do we have right now? Working, breathing, living. That's actually delivering. That's actually delivering.
• 21:41
  Speaker B: We've helped a company, for example, to launch three AI products. So it was an AI transformation case. And by the way, one of the pieces which is like I could illustrate through this use case is also again delivering like collecting data from those conversations and then delivering it to the C level to inform them on what's happening. And so in their case as well, we helped to educate their people in AI. We helped them to think about potential new product ideas, do some sort of MVPs and then kind of distill that to the C level and say, hey, here's what's happening. And then the C level went in and empowered certain groups and saying like, oh, this is actually a super cool idea. Like we actually think this makes strategic sense. And then they ended up launching three AI products in the span of six months. And it's a public company, so they don't want to really reveal everything that the breakdown of their revenue, but it made them quite a few millions. Now that's one example which was very, very tame.
• 22:42
  Speaker A: And let me explain any product. Can you just describe that for a minute? Like, is this like.
• 22:46
  Speaker B: Yep.
• 22:47
  Speaker A: Because what some people sell as an AI.
• 22:51
  Speaker B: So they have very Little tack before.
• 22:54
  Speaker A: I guess my point is like, we have a lot of CTOs, CIOs, IT directors, VPs of IT being pressured to. We got to go AI. We got to go AI. And every time they have a meeting with their, their senior leadership, it's like, what are we doing with AI? What are we doing with AI? And so that can mean a lot of things. It could mean we custom built.
• 23:10
  Speaker B: Yeah, they done a lot of trainings and, and consulting and things like that. And then now besides those trainings, they've built those like kind of AI follow ups on top of them that help their customers to actually practice what they preached during the trigger. And basically instead of like just doing this one time workshop thing, it's like, yeah, it's this one time workshop thing. But then we also have this like automated but personalized at the same time support for the next six months. So it's very simple. But it, it adds a ton of value to what they do because, yeah, the retention from workshops traditionally is very bad. So that's an example.
• 23:56
  Speaker A: Yeah. Is this like a bot that sits on a desk or something or.
• 24:00
  Speaker B: Okay, yeah.
• 24:02
  Speaker A: What's the most practical way someone could implement AI in their organization right now? Is there anything like a, is there like a, like a pathway or a. From, from creation to execution to go back to the beginning? Is there a place to start that someone should start if they're just getting pressured? Like, we've got to be an AI. We've got to be an AI. Is it? Is it? Why don't we just deliver some really, really good, I don't know, like.
• 24:23
  Speaker B: Yeah, so, I mean, what our. So if you look at the MIT report on AI, they say that like, companies are twice more likely to be successful actually if they buy versus if they build in house. Which does make sense to me because most of them don't have the expertise to build in house. So you can build some small things in house. But yeah. And so our basic AI adoption track, it basically starts with understanding what AI can and cannot do. So it's like, okay, let's develop accurate beliefs about how you should think about this thing. For example, right? Like you could say, okay, suddenly AI is actually really good at reading a bunch of qualitative data, a bunch of text and summarizing it. And it's like, okay, that's good to know. And now you can think about where could you apply that? How does that work? So that's number one. Then we would be like, okay, so let's then identify what are the actual high impact AI opportunities in your workflow. Let's map out what you do every day and are there any pain points? Is there something we could automate or do better, et cetera, and kind of give you some insights on that? Then step three could be looking at AI tools. So there's a lot of AI tools out there, right? And back to like buying versus building in house. Like, sure, you can build something in house, but like, would it be easier to look at like what's already available? And there are certain things like I was talking to our bots, talking about like, okay, we want to like screen our code for vulnerabilities and increase security using AI. And it, it, you know, it recommended me a couple tools and one of them is a public company, the other one is done by Microsoft. It's like, I don't know if I would have wanted to do this in house. I would rather just get a tried and true solution.
• 26:17
  Speaker A: So it needs AI to buy AI first of all. Use AI to buy AI. Crazy.
• 26:26
  Speaker B: All the way down, ironically. Yes. But our tool is very specialized, so it's not just like, like this ChatGPT thing. It helps you specifically with this thing. So I think that's like someone has to build.
• 26:38
  Speaker A: Someone has to build.
• 26:39
  Speaker B: Yep. Yeah, yeah, exactly, exactly. And then I think what's also useful is like, you know, understanding what's, what's kind of your vision. Like how does this apply with your personal, like organizational goals, personal values, what's your like 6, 30, 60, 90 day plan to make this real? So we're also helping people with that, like taking first steps. And then. Yeah, so this is some examples, right, of types of like, we call them, like exercises that can be very useful for people who are going through AI transformation. And then imagine you sort of like aggregate all that data and you have like hundreds and thousands of people using it and then you aggregate that data to the C level and you're like, hey, people identifying this kind of issues in their workflows. And this is where AI seems to be most useful potentially. And this is some MVPs that they've built, but they might need resources if you want them to go further with this. And these are things you might want to empower because they seem to be kind of like they got the most traction just from building some MVPs. And they seem to be most frequently mentioned by the people in the company as like biggest challenges.
• 28:05
  Speaker A: Fascinating. With this idea of getting to kind of like the root of the real problems around change. And so there has to be some guiding principles. There has to be some guiding principles even behind any software or artificial intelligence. What are these guiding principles, in your opinion? Or what's your.
• 28:28
  Speaker B: Yeah, there's a bunch. But basically one of the big things is that like change is constant. And rather than sort of viewing this as a project, what we tell our clients is that like, it's like everything is always changing and it should be more like you building your adaptability muscle. And by you, that basically means every employee in the organization, there should constantly be sort of an awareness of what's going on, where are we going, what does that mean for us, what kind of experiments can we run, and then evaluating those experiments and then scaling successful ones. So I think one of our favorite. I just talked actually to the author of the book Lean Change Management, Jason Little. Talk to him before talking to you. And yeah, so that's kind of his philosophy as well, like that. It's, you know, you sort of approach it in the same way you would approach sort of any kind of lean process where you sort of run experiments, you measure, you learn, you repeat. And it's not kind of this like, even if we look at like the ERP project, like, right, so first of all, you would try to learn what are the different approaches that work, right? So okay, you tried that training program, like you measure the effectiveness, like, is it working? And if not, why not? If you kind of use our tool, then that's embedded there because we're constantly collecting the data on what is actually, what are the challenges, what's working, what's not working. And then even after you implemented it, right, you need to still be aware of like, okay, does it actually serve the needs and how can we improve this further? And potentially, if you keep improving that over time, then you don't need to do this kind of massive 10 year or 5 year projects. But with certain things, you can just like constantly have kind of things ongoing. Like with AI transformation, for example, I don't think AI transformation is a project. I think maybe inner ERP would be a bit closer to a project if you're just like, okay, we bought this SAP thing, we just need to install it. But like most things like innovation, AI transformation, et cetera, it's not a project. It's sort of almost like a mindset. It's a process. That's.
• 31:10
  Speaker A: It. Leaders call their end users customers. At least they should be calling them customers. So this is like ongoing customer engagement.
• 31:17
  Speaker B: Yep, yep, absolutely, absolutely. And you know, as if you stop innovation, if you stop AI transformation, That's basically that. That's the time, the moment you start dying. It's not a good place to be.
• 31:34
  Speaker A: How do we make people. Well, I guess we probably could. We probably could make AI agents that actually people like more than people. But how do we. How do we bring back that. I'm sorry, saying bring back, like, it's already gone. But how do we make it have a personal feel? Like people want to engage with this, so they don't think it's just. It has to be done right. But how do we. Yeah, how do we differentiate? Kind of like in the old stupid bot versus, you know, like an agentic AI type of experience?
• 32:09
  Speaker B: I think there's two things, and it's back to the things we've discussed a bit earlier. So number one is, I think a lot of it is about what does the bot know about you? And, like, the memory and the context engineering. It should be smart. That's the answer to how do we not have a dumb bot? I mean, that's a lot of engineering and ensuring that it remembers the right thing about you at the right time and that it knows your context, it knows your role, it knows your struggles, it knows your values, it knows what those discussed in the past. And it takes all of these things into account. So that's number one. But number two is in terms of also the human touch specifically. I think that the bot could help with that too. Right. So back to what I was describing was like, connecting people. Right now, people are not connecting. We're just specific organizations. But using those AI agents, you could actually have that effect where suddenly the right people can find each other and be like, hey, I worked on this thing. Here's the book that helped me. I recommend you to just check it out. Or like, hey, I'm working on this problem. Seems like you're working on it too. Let's meet for coffee and brainstorm how we could maybe help each other or how we could solve this together. So, yeah, there's a couple of answers that come to mind.
• 33:32
  Speaker A: Of the. Okay, do you have any favorites? Do you have a favorite LLM or does it change every week? I'm just curious, like, do you have any favorites? Is there anything out there that we need to, like, request ahead? I'm a Claude. I'm definitely a Claude guy. I've gone back and forth. I. I liked chat GBT for a period now. I really, really almost despise it. I feel like it's like, really. I don't feel like it's learned or responded to me or learned me? I feel like it's more. I, I honestly, I just, I don't know. People throw the term woke around a lot, but I feel like I just can't, like, I'm not really, I don't really get the. I don't really get what I need sometimes. And there's times where it really knows my voice and it pulls stuff out. I'm like, wow, how did it know to feed me this, that? And I felt like it knew to feed me that because I fed it.
• 34:22
  Speaker B: Yep.
• 34:23
  Speaker A: So I, I know that sounds stupid, but can it feed me anything that's like I, I need it to feed me stuff that I would have thought of, but I didn't.
• 34:32
  Speaker B: That's what we do. And the way we do that is. And most LLMs, right, they are kind of pandering to you. And it hasn't, but it has a lot of disadvantages too. And so that is why we have kind of experts that kind of tell it what to do. So rather than just like, oh, Phil has an AI problem, let's just tell AI to solve his problem. It's like, no, use this framework. This is tried and tested. Run him through these questions, help him understand whether this is happening or that is happening, and then shift towards an action plan. So we sort of, we instructed very specifically what to do. And that has that advantage of like actually uncovering your blind spots basically through, through two things. One is the reflection, right? So we kind of, we, we ask you questions rather than sort of just telling you the answer which might or might not be true. And then second of all is by giving you frameworks that maybe you've never heard of about even, and the AI would not pick up because, you know, it's, it's trained on the Internet and on the Internet, sometimes it's hard to say what is real and what is not. Like, you know, that's where the value of real expertise actually shines.
• 35:52
  Speaker A: Now if we took these questions and frameworks and threw them into. Are we taking these questions and frameworks and throwing them into any various different LLMs and are you getting different answers? So like, okay, so my, my partnership company, so App Direct, we, we purchased devs AI. So the devs AI guys bought us the massive aggregator. As far as I know, we're the only enterprise, like SoC2 compliant, whatever, 2700 GDPR compliant AI aggregator with all 50 plus some odd LLMs all aggregated in there along with, you know, private servers and being able to promise an enterprise company that these LLMs are not learning on your data in paperwork. So why. And we're able to, if you're smart enough, like you said, you don't want them to build it yourself, but you could have, you can hire anyone else to build your agents for you. And I don't know if that's considering buying versus building. It's kind of like outsourcing your build. Is that a buy or is that a build? I don't know.
• 36:57
  Speaker B: That's kind of, I think.
• 36:59
  Speaker A: How do you, how do you feel about aggregating multiple LLMs to into one agent? Gro Claude chatgpt all some.
• 37:08
  Speaker B: What are you trying to do? Right. So if you're trying to like do strategy creation, for example, coming back to our topics, then you typically can wait. So it's, it's fine. You can like aggregate them. You can like compare different, different ones. For us, we sort of had to pick the best one because we have instant communication for the, for the AI coach. Right. So people don't want to wait 10 minutes for one reply. And so for us it's like we, we have backups. So it's like if one is offline, then we use another one, things like that. But fundamentally we, we just use primarily cloud these days as kind of a primary one. But for some use cases like on the strategy creation, right, where. Yeah, like, for sure, like you, you definitely want to. And I know some friends who use Perplexity for that reason because you can easily like pick like the model there and like have different tabs and ask them the same thing and just like have different tabs there and not have to go to like five different tools and pay for five tools. You just pay for one on basically.
• 38:15
  Speaker A: So how do you get around dealing with getting different answers? Do we just know like, hey look, you know, 40% of the time there's going to be a fail and people just need to know that.
• 38:27
  Speaker B: I mean, you mean with our tool or hallucination.
• 38:30
  Speaker A: Hallucinations results, guiding people down the wrong path? Hey, we just guided 40% of the end users down the wrong path. Are we connected? I mean, maybe we made the wrong choice or something. I don't, you know, I mean, I'm, I'm assuming you get different answers people get. You're going to get different answers. I mean, that's.
• 38:48
  Speaker B: Yeah, yeah, we actually. So our answer has been creating this very constrained experiences. So rather than just like, okay, you know, like we have the whole Internet here, we are like, no, this is specifically an exercise to do this one thing and do it well, and that tends to produce very good results. There's very little hallucinations because we specifically tell it what to do, basically. And the models are already good enough where they don't suddenly start talking about the weather if you set the parameters strong enough. And that's been kind of our approach. And then when we have those constraint parameters, then the question becomes, okay, how do we forward the person to the right content in the right time? And so there we have a recommendation agent that also works quite well. So basically it just tells you that, hey, you know, considering your problem, this might be the exercises that might be helpful for you. And that so far worked super well for us. But that's our approach and it's kind of the approach of like going very specific versus very general. So I think the Claude has an advantage of being very gentle, but also being a disadvantage of like, potentially insinuating a lot and like, you know, almost knowing too much.
• 40:06
  Speaker A: More constraints, more constraints. Baby steps, baby steps. From. From agent to agent to agent.
• 40:12
  Speaker B: Yep.
• 40:13
  Speaker A: Super cool. Okay, where are we going? Where are we going in the future here? What's going to happen is, are we. Are we in a bubble? Are we in a bubble? Are we in a bubble?
• 40:26
  Speaker B: Define a bubble.
• 40:27
  Speaker A: I mean, massive amounts of money being invested into a new technology with. Yeah. Only actually producing 12% of the revenue that's actually being invested into it. So, I don't know, simplest way to say it, $100 billion invested in AI and it's only producing 10 billion. That's not the real numbers, but it's probably not too far off.
• 40:50
  Speaker B: I think the worst case scenario will be the Internet bubble type scenario.
• 40:55
  Speaker A: Yes.
• 40:56
  Speaker B: And what I mean by that is.
• 40:58
  Speaker A: Will it really be bad? Will it really be bad or will it just be some standing heroes?
• 41:03
  Speaker B: Yeah, yeah, yeah. I mean, I mean, a bunch of companies will definitely go bankrupt, but they always do. So. Come on. I don't know. Hardly. Yeah. Because I mean, to me it's like, I don't know. I haven't lived through the Internet bubble. Right. But it's like looking at it now, I'm like, the Internet took over the world. Like, it wasn't the bubble exactly.
• 41:26
  Speaker A: Like, I lived through it. I lived through. I thought it was pretty freaking cool. I mean, I don't know. I mean, everyone else was. Oh, what are you talking about, Phil, you ignoramus? But no, I mean, yeah, I mean, you kind of see it.
• 41:36
  Speaker B: Investors would stop investing for a little bit, and then companies will. The main comp. Like, companies will grow a bit. Slower and they would start just like being more profitable from earlier days. And yeah, so I think that is possible. I think it is also possible that like that we will sustain the bull run actually because it could be that the gains would be so big that, yeah, that basic.
• 42:04
  Speaker A: Or be smart and know what you're seeing, like kind of be more, I think we're probably more aware of yet people are still, there's still this kind of frenzy going on. Okay, so that's the Internet bubble question. Well, how do you survive that then? Build good stuff, get acquired basically.
• 42:28
  Speaker B: I mean the consumers are still going to be buying stuff, right? It's just more about like the investors might be scared and there might be like a shift in the markets. And then what you do is you just like you rely less on investors. That's what you do.
• 42:44
  Speaker A: How about this? What's your advice to C level technology leaders? CTO, CIOs dealing with C. What's your advice to the CTOs and CEOs? The CEO is being pressured to, I don't want to say lie maybe over embellish on their level of AI integration in their companies. And what do you say to the CTOs being pressured to actually implement AI? What, what's the advice? Because a lot of people kind of just like staring around like, yeah, we're thinking of buying, we need to do this. And they're, they're kind of looking for the low hanging fruit. And I think you kind of mentioned at the beginning like, hey, we have got this thing that needs to analyze. We get these, I don't know, government contracts that are 30,000 pages long. And we need to know from a construction standpoint how many Phillips head screwdrivers we need and how many bolts and how many toilets got delivered and how many toilet seats. And we used to do this kind of manually and now AI can do that for us.
• 43:40
  Speaker B: I don't know. Yeah, I'll need to go in a few minutes. But basically I would say that, you know, a lot of it is the quality of the strategy which comes from like how much does the CEO actually understand AI. And so I think step one is educate yourself and really like spend a lot of time on that and try to understand what's really happening. How does the technology work even if you're not just, it's not that scary, like dive into it, like understand what it's doing.
• 44:11
  Speaker A: CEOs educate yourself on technology, that's.
• 44:14
  Speaker B: And then once you understand where your industry is moving, where the world is moving, where you want to move as part of that, that's your strategy, right? And then it's strategy execution, then it's change management.
• 44:31
  Speaker A: Thank you very much. Been absolutely a pleasure having you on. Can you help me pronounce your last name again? I didn't even. We didn't even do this at the beginning because we just jumped on so quickly. But Dima.
• 44:40
  Speaker B: Sorry, was. Yeah, Dima said Altkin.
• 44:41
  Speaker A: Yeah, that's not too bad. I can do that. I can bang that one out. Final last words. Final last words to any listeners out there. And sea level IT managers, IT leaders.
• 44:53
  Speaker B: Well, if you think about change management, you definitely need to hire McKinsey. You're just being okay. Yeah, no, I think like, again, I think like what we just spoke about, I think fundamentally actually beyond even change management, like understand what's going on, like, learn what this is and how this will affect you. Because I think like, people are either overly, like ambitious or they are actually underhyped, you know, about, about the, the, the change that's coming, I think it's going to be massive.
• 45:30
  Speaker A: So it's gotta be somewhere. You're saying somewhere down the middle. You gotta take the middle path, as always.
• 45:35
  Speaker B: Yeah, but the middle path, I think, is more ambitious than most people realize. Like, I think, I think the middle path is something like Klarna that just said that they fired 60% of their people and they became profitable and the revenue grew. So that might be the middle path. And I know that most of it is bullshit when they say, oh, we fired this many people. But for example, with, with Klarna, it seems real.
• 45:59
  Speaker A: What if they just needed to find. What if they just wanted to. What if they just wanted to signal to the marketplace investor growth. So we're just going to lay off a bunch of people, regardless whether it had anything to do with AI. We're going to lay off a bunch of people. It's going to signal we're going to buy back stock. When do a stock buyback, it's going to signal growth and it's going to increase stuck. And then the CEO is going to quit and say, see you later, and then they're screwed. Is it that? But how is it real?
• 46:21
  Speaker B: Well, but I think what makes this different is the fact that, like, the job seems to still be done. So rather than just we fired the people and like, that's it, they actually have agents that are doing customer service and they're doing marketing and they. A lot of inefficiencies that they. Yeah, they automate it.
• 46:43
  Speaker A: Awesome. Have a wonderful rest of your day. And thank you so much for being on Dissecting Popularity Nerds.
• 46:48
  Speaker B: Yeah, thanks so much.