Tiankai Feng on why your AI strategy is failing (and what to do about it)

Speaker A: Yeah, I'm doing pretty good. How are you doing?

Speaker B: Yeah, I'm doing really well. It's really exhausting, but also really nice at the same time towards, uh, end of the year. I feel like I've been at so many conferences and talks this year, which is really nice and was really rewarding,

Speaker A: but so many performances.

Speaker B: Yeah, that's true. Also singing and playing the piano. Right. But now it feels like, yeah, I can take a break. It's fine now. Like, I need that break now.

Speaker A: We should start a new kind of show called Data, or AI. Uh, has, um, talent. Something like that.

Speaker B: Oh, I would love that.

Speaker A: Yeah.

Speaker B: Oh, you and I can be the judges. Right? And then we go.

Speaker A: Can I. Can I get one of those red buttons where I just go, like.

Speaker B: Yeah, exactly.

Speaker A: We should. Yeah, we should do something like that.

Speaker B: And you have a really nice glowing logo now in the background. That's really cool. Awesome.

Speaker A: I had to invest in that. It's, uh, I think about a year now. And, uh. But that's cool. But I was. I have this cast on now. It's like putting on sweaters is really difficult, so I thought I had to put on something jovial because I'm talking to you. So I put on my. My summer shirt right here. It's like, two degrees outside, but I thought I need something jovial.

Speaker B: That's so nice. Oh, thank you. No, I mean, it does give, like, a new summer vibe, which is really cool right now.

Speaker A: Yeah, exactly. Right? Yeah. It makes you feel warmer even though your arms are cold. But what a year it's been for you, though. A new book. And I guess you're doing a publication tour as well, or promotion tour, kind of.

Speaker B: I, uh, feel like it has been a publication tour anyway. Like, I mean, my talks have been evolving now from m. Humanizing data only to humanizing data and AI. And some people only want to hear about the AI part. So it's kind of like, um, a mix and match of a repertoire of content I have now that I present. But it works, right? I think, like, the human centricity of it all, even if it's. It's almost ironic because the reason why I get invited so often now, it's because no one else talks about it. So they feel like someone talks about the human side of it. Let's bring that person in, because that's a refreshing take on it. Because everyone else talks about architecture and the kind of AI technologies and model comparisons and all these kind of things, but I bring in a different part of view to it. Right. And I think that's really cool. I hope it makes a change and an impact somehow. Right. And to have a little bit of a mindset change about the whole thing.

Speaker A: No, absolutely no. It's been very inspiring thinking about the human aspect of it, because I think you're absolutely right. A lot of people do talk about tech, talk about the benefits we can cut, so become so much more proficient and efficient, but nobody really talks about what needs to change internally with people. And I think your first book really touched upon that. Did you, did you see this funny image? And I think I'll put a link in it once I published this was, uh, the guys from Dojo, AI, Duarte and Luke. They were at a conference in Amsterdam and apparently there were like two stages. There was the AI stage and the non AI stage. And they were presenting at the AI stage. And the non AI stage was empty. It was like three rows of chairs and the AI moment was packed. I was like, just like you said, everyone wants to talk about this AI stuff.

Speaker B: Oh, that's funny. But also so sad for the non AI stage. Right. Nobody going there is really sad.

Speaker A: Can you imagine? Yeah.

Speaker B: Oh my God.

Speaker A: I want to talk about how we can help people work with technology. And there's like no one there.

Speaker B: Exactly.

Speaker A: Been there, done that. Hey, but last time we talked about humanizing data strategy, when we talk about AI strategy, what needs humanizing in that part? It's mostly built around humans, right? Human input, human knowledge that's been gathered and parsed and made available. What needs humanizing around that?

Speaker B: I mean, I think it comes down to the five Cs again, right? Which are the same five Cs as in data strategy, but it has a slightly different meaning or even expansion in most of the cases that I'm trying to address. So for example, with competence, right? Yeah, it's a lot more about not how you want to use AI more, because most people use it in any kind of extent already and it's using it more wisely. But using it more wisely means also to know what you should use it for, what you should delegate to it, but what you should also keep as a human only task that you shouldn't delegate. But knowing the differences means also you need to understand how the technology works, at least conceptually, so you don't make kind of the same mistakes. And I think with especially generative AI, it kind of made it into the mainstream. Right. And now it's really, uh, there. But that's just one of the reasons also why competence, uh, has, like, this different meaning. And just one other example from one of the other C's would be communication. Right? Where in data strategy, it was all about human to human communication. Like, how do we talk about data? How do we human. That basically creates a value around data and communicate in the right way. But in AI, it's really about human to machine, machine to human, and human to human and machine to machine communication. Right. And that has a very different meaning. So. So the question is, how do we put it all together?

Speaker A: Right?

Speaker B: Because there's a lot of different combinations of how communication works. What should we look into and how should we address in the right way? So, yeah, those were the kind of things that led me to write the second book.

Speaker A: Yeah. That's interesting, because the last time you said that, a lot of failures in data strategy can be traced back to human behavior. I mean, but from what you just said, what's the AI equivalent of that? Is it still humans?

Speaker B: Yes, I think so. I mean, uh, if we say it in the different way, right. AI is always being trained with, uh, data. And the data itself, first of all, is always biased, Right? Because the more we use historical data, the more we dig into the past. And let's be honest, human beings weren't always great in the past. Right. We did a lot of mistakes. So the further we go back, the more mistakes they'll make. Exactly. The more we actually get AI learn from them. And the thing about it is, AI doesn't judge if it's right or wrong. Right. It just knows I need to learn from this. So in the worst case, it actually amplifies and accelerates the mistakes from the past into a direction how you cannot stop it or you notice only far too late. And so making the choice, first of all, having done something in the past that was bad and knowing now that it's good, that is very human. But having to teach AI, or making the conscious choice to let AI only learn from the good parts and the right parts of it and not from the wrong parts is a conscious choice that has consequences on how we do it. Right. And then, of course, we have the whole thing about using AI for good versus for bad. Right? And sometimes the bad is not even a conscious or intentional bad. It's just something that we haven't considered as a consequence because we haven't, um, talked to the right people about what might go wrong. So we kind of let it out there and it starts happening, doing bad things. So I think it still is a human being. Right?

Speaker A: I think that's the vending machine example where wanted to reach out to the FBI because of a, uh, detected fraud or something.

Speaker B: Oh, my God. I didn't know that. But.

Speaker A: Okay, I think there was this use case. I might have it wrong, but if my memory. I mean, uh, my memory is not what it was 30, uh, years ago. But I believe there was this case that they were testing Claude in a kind of vending machine scenario that it would run as a vending machine. And I think it ended up, uh, wanting to contact the FBI for some reason. So like you said.

Speaker B: Oh, my God. Okay.

Speaker A: If you, if you want to use it for, for, you know, to be bad, I think like you said, sometimes you need to be in there as a human to prevent it from going bad. Because, like, yeah, it doesn't have the. It's unable to properly determine what it's doing, and the effects of that is going to be. But do you feel that, like, I mean, we're going pretty fast now, but do you believe that the hype is kind of over, or is there some realization happening among practitioners about the true extent to which AI can really help them?

Speaker B: Yeah, I think we are on the verge of it. I mean, from many things that we just see in the news. Right.

Speaker A: Yeah.

Speaker B: It does feel like the bubble is about to burst. Right. And we see that already with Nvidia not really doing that well anymore. And I feel like also OpenAI, for example, is not as in the good light as it was two years ago anymore. Right. So all of these things, I think, are really strong indicators that something's going to happen. But however, I think still AI is going to stay.

Speaker A: Hold on, let me dump some stock here real quick. I need. I don't know what inside information you

Speaker B: have, but, uh, he's like, oh, sell, sell. Exactly.

Speaker A: I wish I had stocks. But anyway.

Speaker B: No, but what I'm thinking is let's say the bubble is about to burst.

Speaker A: Yeah.

Speaker B: It's still going to be a, uh, change forever. Because AI is not going to just disappear. Right. It's more about what is remaining. And I think what's going to remain is the technology being used in a more balanced extent. Right. And what I mean by that is potentially that there are not so many big players anymore that are dominating and kind of the whole space. But I hope that it's more the smaller players that are, uh, now find a place. It's a little bit like the modern data stack suddenly becoming bigger and replacing all of the big players. And I hope that's going to happen too. Or I feel like this. And I mean, just looking at also how a lot of people can now get their own GPUs. They create their own smart small language models, right?

Speaker A: Yeah.

Speaker B: They run locally their own little, um, AI kind of models. I think that might be the way to go. Right. Because the more we can specialize and have control over what we want to do, I think the more human centric we can also be. And it's not about then depending on someone else's interest, but more depending on our own interests. And I hope that is kind of how we can use AI in the future as well.

Speaker A: Yeah, it's funny that. I mean, that's an interesting point because

Speaker B: if I look back at.

Speaker A: Because I'm really looking forward to the State of Martech Report 2026, I believe that's coming out. They do it in December 25th, but it's for 26. Um, but if we look at last year's there was a lot of talk about micro SaaS. And uh, personally I've been experimenting a lot with the, uh, likes of Claude Code, uh, you know, building, building micro SaaS solutions for clients. And I think you're right in saying that AI, even though you have the big players, is opening up a market for these smallers. Um, you can use the big players to teach you how to, and actually help you build, uh, local models or models that are tailored to your own personal needs and reducing the dependencies on the bigger ones. But how do you, I mean, even then there's a lot of responsibility that a lot of companies, uh, kind of overlook. And how do you coach leaders right now? Or how do you talk to leaders who are confident in the hype that AI is the way forward, that it needs to be incorporated, but they're terrified of the responsibility because, let's face it, sometimes it is a black box. Sometimes you do need to hold AI's hand and say, hey, check my code for any embedded environment variables. That, or, you know, API keys and anything like that, make sure they're moved to environment variables. If you don't have the basic knowledge of that.

Speaker B: Right.

Speaker A: You're lacking in responsibility. But how do, how do you help coach leaders in this?

Speaker B: That's a very interesting question. I would actually say that what I'm more observing is not a lack of, or the fear of responsibility, it's the ignorance of it almost. Right.

Speaker A: It's like, I don't know what's worse.

Speaker B: Yeah, I don't know what's worse either, but I feel like it's more. It's just not conscious for them. It's more like everyone's doing it, let's do it too. It cannot be that bad. Right? That's kind of the attitude. So they run into it and not considering actually what could go wrong or everything that is really bad. And then bad things happen. Right. Basically they would vibe code stuff, for example, and then suddenly they're being exposed to have a lot of data leaks and um, insecure kind of data storages and then they immediately get hit with the shitstorm and everything is over. Right. Or I mean some other companies or clients just tell them, no, this is not possible. What are you doing? This is not what, what I want to be part of. And you get the kind of feedback. So it's more actually overstepping, uh, almost first and then having to go back to like, oh, okay, then what should I do? And this is where I can usually come in. Right. So this is like, okay, they have now learned a lesson, but now they want to do it right, but they don't know how to look into how to do it right. And that often ends up with me talking about what AI governance actually means. For example, right?

Speaker A: So OpenAI is going to be reaching out to you soon and Nvidia saying, hey, we've lost 60% in stock price, let's call Chiang Kai from all the people.

Speaker B: Exactly. That would reach out to me. No, but I think this is then where I usually try to have like a conversation about AI, ah, governance with people. What does AI governance actually mean? And it usually boils down in a very simple concept into three layers of governance. Right? So you have the data governance of the data that is being fed to it or that's being used for rag, for example. You have then the model governance of it. All right, so the AI model itself and you have the interface governance of where the model is being applied to. Right. And how you basically open it up for people to have access to it.

Speaker A: Uh, oh, that's interesting.

Speaker B: And the thing is every layer has usually its own people that is taking care of it. And they very rarely talk to each other because they in the past didn't have to talk to each other. Right? Or like before this all happened, now it wasn't that big of a deal. So you had like UX designers, for example, or software robots creating software for the interface. You had like data scientists creating really cool models on their own, right. And only using it themselves. And you had data, uh, basically doing it for recording and whatever. Right. That's not really for AI applications In the past so much and now it all comes together and everyone expects it to go really clearly. And so this is the first lesson I try to teach decision makers and leaders to think about the three layers of where they sit in their organization, how they are working with those, uh, teams in those three layers and how we can bring them together to have a shared purpose. To say, okay, now we have this use case that go through all of these three layers. Let's make sure it all works together and not just throw it over the fence and then everything will be fine. Right. It's really about finding the way together. And yeah, that I think is a good, great starting point.

Speaker A: Oh no, absolutely. I mean, I think that was definitely part of your first book, is that we need to build those relationships and kind of bring everyone to work together. And with those three governance layers, I mean, do you feel. I think there was actually a research on this as well, the Dunnar Kruger effect of AI making people overconfident and how giving them more confidence that might be healthy for them in the positions that they hold.

Speaker B: Absolutely.

Speaker A: I mean, AI systems are basically improv actors. I mean.

Speaker B: Yes.

Speaker A: And yeah, yeah, I mean, well, uh, uh, true story here. I, I did something in Claude and Claude has memory all of a sudden. And right now every time that I talk to Claude and I say, hey, I need to, I need, I need a little script for this or I need a, I need you to review a script for that.

Speaker B: Right.

Speaker A: Usually when I'm working, anything, uh, with code I do with Claude and it says, oh yeah, no, that's, that's that problem, man. I can tell you stories about that. I was like, this is an AI system. It's wasting water on telling me these light hearted improv jokes around the work that I'm doing.

Speaker B: Right.

Speaker A: I can't trace back to a discussion that I had with Claude where I asked it to kind of be a little bit salty and uh, you know, a little, a tone of, tone of humor. So I have no idea where he got that. I lost track of the question I want to ask. Yeah, yeah, we were talking about improv actors, but.

Speaker B: Yeah, how do you.

Speaker A: Yeah, so you have the three layers of governance that now use AI in their own way, making them feel confident. But how do you, I mean, what's your experience so far in building the trust between these three layers of governance then? Because they're, they all feel overconfident. They all feel that they're right because they have this AI agent telling them so. Right.

Speaker B: Oh, okay. I, uh, See what they're saying.

Speaker A: How, how would you break that down?

Speaker B: Yeah, and that's a good point.

Speaker A: Is there an, is there an overarching governance that says, hey listen, we need to keep realizing we're just talking to systems on, you know, probability systems? Mhm. Yeah, that's. Yeah, I can't even answer that one for myself. No.

Speaker B: So you're absolutely right. I think what was interesting about your question is it implies that all three layers are using AI also to do the governance, which is not something I actually haven't thought about that much yet. But it's a really good point. But what happens is, right, that for all three governance layers they do have customers. Right. If you so want to say in a very simplified way. Because they're governing it not just to govern it.

Speaker A: Right.

Speaker B: Uh, they govern it to force someone to use it. Right. That is the whole point of it all. The question is only how can I prioritize also the AI world and all of the things around it as my customers? So that might be right, that um, my data needs to be good enough for a data scientist to use it to train their model. Right? Yeah. But it might also be that my data needs to be good enough so an AI agent can get the right context from it. So it then actually produces the right tasks and the results for it to done. So what changes mainly is my customers are not human beings only anymore. Sometimes they're automated bots, sometimes they're agents, whatever they are. Suddenly the world of who's using the things I'm working on is changing drastically. And then you have agents to agents usage, they coordinate together now and multi agent all of this stuff that's going on. Ah, so it's a little bit where, how do we as human beings deal with it. Right. Because we cannot let them run completely autonomously either. Right. And that's why we talk about human loop often. So where, how do we make sure that it's both machine ready to be used by something that is not a human, but also it's usable by a human being? Because I need them to trust me as well, otherwise they're not going to send that agent to me. Right. So I think those are the different layers and it goes back to um, the general collaboration principles. Like in the first book that I wrote I talk about co creation instead of transaction. Right. I think that's still very much there. Let's deal with it as a. We are building something together and not as I serve as you, you serve as me and learning together. Exactly.

Speaker A: Yeah, yeah. That, that's, that's definitely a good point because uh, uh, you know, if you learn in isolation, I think it's, it's the skills that are like. I think you also made a point about that in your book around uh, involving HR more into the hiring process around data or making sure that they know what to look for, that even they are educated or literate around data. And now AI, of course, to make sure when they go through the hiring process they're not, you know, you know, what's the word? Side.

Speaker B: Um, sidetracked maybe.

Speaker A: Not sidetracked. Hit from the side. There's a right.

Speaker B: Got it.

Speaker A: Yeah, yeah. They're not hit from the side by blindsided. Sorry. Yeah, by false bravado and CVS saying, hey, because you really need to test for this stuff like you said. Because in effect if you hire someone who turns out doesn't have the skills that they thought they had, you know, the false confidence there.

Speaker B: Right.

Speaker A: Erodes trust within your organization. But there was another, I love these acronyms in your book. You had the five C's and the other one was tag. I don't know, I don't. I want to ask. Do you remember that one? But it's your book. Of course. I need another coffee. But how does tag. So for the people listening, tag, uh, was the acronym Chiang uh Kai used for talk, act and guide. How does that now apply within AI? Because it's a, it's a leadership, a kind of top down approach. Does HM, that change with AI?

Speaker B: Uh, yeah, absolutely. And I think why it changed with AI is that my feeling is that a lot of the leaders that are trying to force AI innovation in their organizations are not using it that much themselves yet to the extent that they think the organization should do it. Right.

Speaker A: Is that the same with data? At the time you had a lot of data leaders. Who is it? Because a lack of hands on work.

Speaker B: Exactly, exactly. So it's like a CEO saying we need to be data driven, but they never look at one report themselves ever. Right. But everyone should be.

Speaker A: They should. And uh, they do. They do.

Speaker B: Exactly. So it's probably the same logic with AI where shareholders ask them for example, to be more AI driven. So they're going to tell the whole organization we're going to be more AI driven now. But they don't use any AI in their own. So it's for them just an abstract concept where they think they know about it theoretically, but without the practical experience of at least being an uh, AI user to some degree. Right. It doesn't really work that way. So I think this is a starting point also about talk, act and guide. Right? Like talk about it. Sure, yeah. It's important to communicate as a priority, but act like it too, Right. You have to somehow be credible. Right. When you talk about the things that you're talking about. And ideally, if you know good enough, you can guide others to do the right things with it too.

Speaker A: Right.

Speaker B: Where, uh, it can just say, now, based on what I know and what I have experienced myself, let's do it this way, not that way. Because if I were a customer of mine, I would like to have it this way. For example. Right. So that's kind of the idea to also apply that same framework there.

Speaker A: Um, here's an ad hoc question then. Have you seen any examples of leadership, leaders, leadership that have forced themselves or invested in themselves to learn about AI and been able to take on that tag role lately?

Speaker B: Yes, absolutely. Yes, I do. Yes. I mean, so, I mean, most of the leaders I talk to, they know it on their own by now, Right. So they kind of feel the need to be there, especially when something bad happened and they feel really lost that they could have not anticipated at all themselves.

Speaker A: Okay.

Speaker B: I think I need to get into it. Right. I cannot. This is. How have I not predicted that this would happen? Right. So they're kind of now being forced to go into it. Um, and others, um, are just feeling they need to keep up with the industry. Right. Um, simply said, they're also afraid they're going to lose their jobs. Right? So especially like middle managers and so on, they feel like if they're not keeping up with the trend of AI and being able to navigate the world with AI, then they're going to be irrelevant in the future. Right. So also, just from a survival instinct point of view, they're trying to learn more about it now. That comes back to that.

Speaker A: I think one of the first quotes that I, that I picked up on when AI started really having impact on productivity, it says, AI is not going to replace you, but someone with AI skills is going to replace you. And I think that. So from what I hear from you, that's still the kind of the point to this day. But here's maybe a strange question. Is AI making people intellectually lazy? Um, what have you seen within the organizations that you worked with? And I'm not talking about the LinkedIn fantasy land. We live on a few, uh, 30 to 60 minutes a day.

Speaker B: Yeah, uh, lazy is a good point, I would call it. So it's a choice first of all. Right, yeah.

Speaker A: Oh, that's. That's a good one.

Speaker B: So it can make you lazy, but I hope that we may. It makes us more productive. That's what I would say. Right.

Speaker A: So. And that. Is that what you've been seeing though, the productivity?

Speaker B: Yes, and I hope that is something that. Because productivity also implies that you're still in the driver's seat and you're not outsourcing their decisions itself, but you are the one who is making the final call of what you're using it for and what the output is that you want to use it for. Right. So that's kind of the idea. But it is true, right, that it can make you lazy. You can. Like the difference might be. Right. And we just talked about LinkedIn AI slob. Right, the bubble, for example. Right. You could also say, help me write a LinkedIn post about whatever is currently hot and it gives you a list of current news and then kind of gives you a text of a post and you post it. Or you could ask it to. Can you help me, um, uh, summarize these five thoughts I had about the dangers of using genai for financial decisions? Um, I had this in mind. And can you summarize it in the right way? Um, I want to make a post about it, which would be productivity reasons. Right. So I have already all of my original thoughts, but now I just want it to be sounding coherent at least. So give me some help at least to put it together and then I can go through it to put my tone of voice on it and then I post it. But that's the main difference, Right? You can either just go broad and do everything with AI leading you towards something, or you stay in the lead and use it for something. And that's the difference between being lazy and being productive, I would say. Right. Where you can, of course, outsource completely your critical thinking and you have only the broadest intentions, or you still have the clearest specific intentions and you use AI exactly for that. And for those, AI is more effective anyway. Right. We know that the more specific you are in prompting, the better you get results. So even that plays more into the functionalities of generative AI Anyway, so, yeah, that's kind of. I hope that the balance that we can strike where it's not lazy anymore, but more productive.

Speaker A: Yeah, no, absolutely. I mean, uh, I play this computer game. Um, well, computer game, it's finally out on the iPhone. I've always wanted to play it called Subnautica. It's like this underwater horror survival Game and I really love it. It's created by a company called Crafton and Krafton recently said, um, that they are replacing, I mean we've heard this story before, replacing a lot of their employees with AI. They're going to be an AI first gaming studio. Yeah. So this brings back, uh, you know, nightmares, uh, from Klarna who reversed Duolingo. They lost massive amounts of stock value, uh, after they announced kind of the same approach. What I mean is that kind of a signal from the consumer side that we are not willing to be led by AI. How should companies, or, uh, have you been in contact with companies who have been considering same type of moves or have they been scared off by this type of news?

Speaker B: So yes, I think I would be lying if no company has talked with me about it. But I would say I'd usually talk to them in a stage where they haven't made the decision yet and they're just asking in confidence. Right. So yeah. Do you think with AI we could let go of how many people could we save a lot of costs on the human labor, for example? Right, yeah. Is it possible? And the question is always what, what is the goal here? So if we can automate a lot of things, let's say, and we can make agents autonomously taking over the actions, then sure, a lot of the tasks that were previously human, repetitive, tedious tasks, are not needed anymore. But instead of then saying then let's have those people leave, couldn't we actually use it to our advantage to make them, for example, product managers or make them, let's say quality assurers of those agents? Right. And whatever these kind of things, they would be the perfect candidate because otherwise you anyway need those roles. And then suddenly you realize the people that all would help you to actually assure that it's of the right quality and not doing anything wrong, they all left, right? Because you actually made them leave. So the, the question always, that I always try to bring it back is not only from a empathy, uh, point of view that I don't want anyone to just lose their job, but also from an organizational fitness point of view instead of only thinking destructive about human beings, why don't we lift up humanity, right? Why don't we then use the skills of the human beings that makes them more human than AI to use it in our advantage. Let's do be even better, right? We can automate our tasks, let's use those human beings to come up with even better ideas, with even better frameworks and processes, and then we are even better than Before. Right. And that makes shareholders happy too.

Speaker A: You're going for the Nobel Peace Prize here.

Speaker B: Yeah, this was my application video.

Speaker A: Exactly. I mean that, I mean, very noble and I actually love it. I like the word, I like what you said. You shouldn't think about it in a destructive manner and. But you should, uh, you know, some of the money that you're putting into. I'm actually, I think I said, um, um, I wrote it a little bit differently in an article I just submitted for martech. Org. Um, is that if you, if you. Instead of keep. Hold on, let me rephrase that. So instead of reinvesting, uh, money, let's say a tool doesn't work or there's another tool that has A.I. uh, and you spend hundreds of thousands on implementing a solution like that into your organization, but you don't take a percentage of that and say, hey, but we also need to retrain our people. In my opinion, you're doing things wrong. And I'm going to take a page out of your book and say these, uh, organizations need to start thinking about how to also uplift, ah, the skills of the people that they have currently and take that along in their strategy to move forward. This is, this reality though, is are companies really looking the same at the technology investments as they are at the human investments?

Speaker B: Yeah, that's a good question. I mean, I do think it. What I think this is also what you're implying. It's a little bit of a cultural question for organizations, right? How did they treat and look at their employees before? And AI only amplifies and accelerates certain views and decisions on that depending on what that was before. Right. So if it was always a more already human centric organization because it really values the expertise and the unique characteristics and attitudes of people, then I think in, even in times of AI, uh, it would not lose that. Right. It would probably just use it to the advantage to do it. But unfortunately, right. The bigger organizations are, the less these cultures exist that way. And it's just too many people for anyone to have that personal feeling about the workhorse. And almost every employee just can very easily be. Just become a number or a row in the database. Right. And it's not really anymore something where you can relate to. So it becomes a numbers game instead of actually a people management game. And this is, I think then what can really go badly. If you have that mindset towards what AI can help you with and you just think of, okay, and then we have, from this many people, we now have only this many People, and that looks good. And we're saving so many costs and percentages and everything is fine in the short term. And yeah, that, that is kind of the question. I think maybe one other point to this only is the question between short term and long term thinking, because I think with. Although we hope that all the senior leaders are thinking long term with AI, there's just too much uncertainty for anyone to be able to, or, uh, capable to think long term. It's just impossible. Right?

Speaker A: Yeah.

Speaker B: Things are changing so quickly that no one actually knows how the technology, how the market, how the industry, how society is going to evolve, that everyone is forced to make short term decisions. And this is why layoffs can be so easy to be decided on because it's like, okay, let's just get it done first and we can always rehire people. But no one thinks about the consequences of that. Right. How much is being lost on the way? Because, uh, just hiring back people, I mean if they want to come back or hire new people, is not going to be the same as having all the institutional knowledge from before for. Right. So I guess that that's the other way of how everyone is acting crazy

Speaker A: or even, even more a darker pattern. And uh, I'm always telling my son, he, he thinks he loves dark comedy, but I've. Even so I need to touch on this dark part bit and this is from experience because even employers don't. And, and cases that I've seen do not want to give them educational budgets because they feel that if you're employing, they're gonna leave.

Speaker B: Right.

Speaker A: That's another.

Speaker B: I think there's this famous quote, right? As in, what if they all learn and then they leave and then the other person says, what if they don't learn and they all stay?

Speaker A: Yeah.

Speaker B: Oh, so then, yeah, right.

Speaker A: Yeah, yeah, yeah.

Speaker B: It's a really good counterpoint to it.

Speaker A: That's a very good one. No, but indeed. But you know, what, what's. And, uh, I've seen that happen. You just request, you know, you go to a course, go to a conference, what are you gonna learn? You. Oh, are you doing this? Even, even with podcasting, people have reached out to me, said, hey, I'd love to do a podcast, but if people associate joining a podcast or going, doing an interview as a signal, especially if they're an employee of a kind of a company, that they're on the lookout for something else. And it, it's a, it's a bit of a strange situation to be in. And I think with AI, um, companies are Struggling to get talent in like the good talent, you know, not the, uh, the vibe. Coding a little brother who, who's making FL apps for, for his quiz on, on Thursday, but people who really have a sound knowledge of how it all works. Yeah, it's very strange, but I think there's definitely an educational, uh, opportunity there. Absolutely. Um, you, you spoke about the, you know, you've been an analyst before. You, you've done tons of work with data and in a previous conversation you kind of call this kind of in some cases one person survival unit. But now with AI, I mean, does that lonely, loneliness kind of the bubble that, you know, as analysts we, we've sometimes sat in, does it get worse or better? Because, you know, do you start relying more on your AI or does it really help you kind of break free and start talking with other people? That's. It's another concern that I have.

Speaker B: Yeah, I think that the main question is what the role of the analyst will look like in the future. Right. So what I'm saying with this is there's a lot of talks of how dashboards will or will not exist anymore in the future. Right. And it might be like anything can be just done through a chatbot. You don't even have to look at a dashboard anymore. You just ask the chatbot. But, um, my point to it, and I think I talked to a few people about that, is that sometimes people just need that visual glance at something, right. And not ask a question every time they want to know something, but they just want to look if it's green or red, green or red, or if it's up or down. Right. And that's the very quick decision. And you don't have to type the whole question in just to get the answer. So I think in that way dashboard will still exist. So the question then is what do data analysts, for example, still do? Right? They will still do the root cause analysis. They will still do it. My feeling is they will also have to take care of a lot of backend stuff. Right. If people want conversational interface in the future, then they need to have the right metadata model behind it and the right data model around the metrics and KPIs. So it works. Right. And this is something that only data analysts so far have been doing. So they need to now make it work. So the chatbot on top actually works. Right. So instead of actually talking more to business stakeholders, they might have to end up doing a little bit more backend work. But nonetheless, it's that whole customer thinking design Thinking of letting your customers or stakeholders be a part of the creation of things I hope will stay right. So there will still be a part that is a lot of based on conversation and communicating to each other and the community part. Exactly. But then some of the tasks will feel more backend and more in the bubble than before. Most likely.

Speaker A: Yeah. So in your mind, what is the most human aspect of AI at the moment? What is the part that AI depends on humans? If you. Okay, let me a little bit of context. You have agents, you have agentic AI, you have a 2A, you have multi agent. There's this whole plethora of acronyms and methods on how to use, use AI. But if we look at, if we take a step back, what is the core human trait that's key to all of that?

Speaker B: I mean, I think for me the very initial thought to the question is we need still human originality and fresh human uh, input for AI. Because otherwise if we stop it, AI is just going to learn from AI generated output again and it's going to be called recursive training. Right. Everything becomes bland and low quality because it doesn't really make sense anymore. It's just all going to be standardized in some kind of weird way. So I think we forget about that a lot, right? Where we are required for AI to stay relevant because otherwise it's going to just do very stupid, uh, stuff in the future because there's no new impulse anymore, no feedback from human beings, no new content from human beings. What is it going to be trained on or being basically fine tuned on. Right. If that doesn't happen. So I think that is for me the most human part of it all. That beyond all of the automation and autonomous actions that AI is going to bring us, we need to be the continuous impulse for AI to actually work. So it's actually depending on us. As much as we think we're going to be dependent on AI, AI is going to still depend on us and we should never forget that.

Speaker A: No, that's true. I mean I've been working with um, notion AI lately as well. But also like I said, Claude, and what I find interesting is that a lot of these tools have this one section, this one field that you fill in that you can explain or kind uh, of define what your expectations are and what it returns. And this feels like that human plug, like, all right, this is where I can plug myself in because if I don't do it, this AI is going to act like the billions of others that are out there right now.

Speaker B: Exactly. Exactly. That's a really, really good point. Uh, it's also right. This is how you make these. A chatbot specific. Like when we talked before about. It makes you overconfident. If you prompted before, as a general rule, to always be critical about what you write. Challenge you, not just challenge you, and not just always agree to everything you do, then it will act like it. Right. But we have to make the conscious choice to do it. Otherwise we're just going to be following the lead again of someone. Right. And, uh, we know a lot of things happen or just for mental health reasons is really bad too.

Speaker A: Yeah, no, definitely. And that's a perfect segue because I had this closing question. It was like, if you would have to remove one human habit that constantly ruins AI strategy, what's the habit you delete from the code base?

Speaker B: Wow, what a question.

Speaker A: Hm. Because we do influence the AI and there's something, some human habit. Is it ego? Is it?

Speaker B: Yeah, I mean, I would have said something like that. Because in the end, the worst part is doing something with bad intentions. Right. And that might be a bias towards specific groups. Right? Like racism, whatever. Yeah, those kind of topics. It might also be to create fake information just to reach your goal and doing it intentionally. All of these things.

Speaker A: Deloitte, Lawyers in America.

Speaker B: Exactly. Oh, my God, what an example. Those are the things.

Speaker A: Sorry, Deloitte. Sorry. I guess I lost a sponsor there.

Speaker B: No, but those are the things that I think ruining AI, Right? Those are the things where, because of those things happening now, we have to put all of the guardrails around it. But everyone is cautious now because everything might be fake, but if we hadn't had those really bad intentions in the first place, we wouldn't have to worry about it. So that's kind of the part where we kind of bit ourselves, right? We're like, well, we made it possible to create all of these very real looking fake things and now we have to all be more cautious. And it's creating like a big, uh, wave of challenges for us. But that's maybe the one habit I would, if I could, I would delete, probably.

Speaker A: Yeah. But against it, it's a great case for, um, embedding ethics and morality into AI, even as a, even as a business owner, to make sure that you stay on track with the identity you have as a human. Not that you want AI to have Chiang Kai before we close. We are doing the same thing as we did last time. I did reach out to a lot of people. I didn't make a name selector this time, but I will do that for all the people who kind of like and, uh, respond to the, to the podcast. But we're giving away five copies of your book. I just want to say if you, if you enjoy this episode, uh, to those who are listening, make sure you give a reaction. If you're watching it on YouTube. I'll make sure to leave a link to the LinkedIn article where you can do that and then we'll draw five names. Chiang Kai will again. I didn't check this with you. We can do the personal message again.

Speaker B: Yeah, absolutely, absolutely.

Speaker A: And then, and then we'll ship out those books to you, uh, as soon as possible.

Speaker B: Perfect.

Speaker A: Thanks again for being on the show. It was great catching up with you. Listening to your. What. What's your take on 2026 going to expand even more? Is it going to implode a little bit or is something new going to come over the horizon?

Speaker B: I think it's going to change. Like as I said, I think it's a little bit about the bubble now being squeezed and we're going to see some smaller players and more a nuanced way of dealing with AI, hopefully in 2026.

Speaker A: I hope so too. Thanks a lot, Chiang Kai.

Speaker B: Cool, thanks.

Speaker A: Bye. Bye.

Speaker B: Bye.