Latest / Beyond Vibe Coding / #S2.17 Don’t Design AI for Your Company. Redesign Your Company for AI - Stefan (Infinite Mind)
Transcript
- Stefan: So I think that's the starting point and that is creating direction. And alignment because if you don't do that, people will fill that void and employees will will fill that void with fear. And investors will fill that void with interpretations that you might not know what you're doing or might interpret it in a different way. So I think this is the starting point on that path to becoming that company. Sebastian: Welcome to another episode of our new season of Beyond Vibecoding partnering with ImpalaSearch. André Neubauer: The go to tech and executive search agency in Germany. Sebastian: In this podcast, we explore the transformational change in software engineering and knowledge work in general. I'm Sebastian Heidemeyer zu Erpen, CTO of Ecosia. André Neubauer: And I'm Andre, CTPO at Trusted Shops. Great to have you back. Today it's another perspective again. We are learning from the experience of an AI Boutique consulting company. Sebastian: Yes, Stefan from Infinite Mind is joining us to talk about their approach to helping companies navigate the AI transformation. It was a lively and quite interesting discussion, and yes, we hope you enjoyed as much as we did. Welcome to our podcast, Stefan. We're happy to have you. And as usual, we'd also like you to introduce yourself to our listeners. Stefan: Yeah, so hi Sebastian, hi Andre. thanks for having me. It's really a pleasure. well, what can I tell you? I've been I've been working pretty much all my professional life at the edge of yeah, technology and how technology actually changes the the status quo. Starting from my my university years building my first software company, then spending some time in logistics, And spending also time in consulting, primarily working with large corporations, looking at yeah, how to actually change the way how they work with technology. and spending quantitative time in the media space, also as CPO at Proceben Sat1, and with media being certainly one of the industries that really had to transform early on due to technology. So that's pretty much my my sweet spots where I've been working in understanding how to change, how to use new technologies, how to deal with innovation. So that's pretty much my sweet spot and combined with with software development and and and digital. So that's pretty much my my professional sweet spot and to be honest, I I need to make one confession. I in that time I've never really or I've seen very few strategies come through live. And not because they were bad strategies, but just because they had to bridge this gap between the boardroom and the execution and the users and the reality. So this is this is the kind of area that I've been working in. And from that I actually founded Infinite Mind together with Robert. now one and a half years ago. And yeah, Infinite Mind is pretty much a boutique consultancy here based out of Munich. And we help companies actually go through this through this change and that is needed to really use technology and AI specifically in a way that is adding value to your business. So we and we do that quite holistically. So we look at Not only at the technology, and I think this is pretty much what we want to talk about today, that it's not just about technology. It's it's usually technology is just 20% of what we do. So it's about looking at processes, looking at how roles and organizations change, taking the people on board and and and facilitating that change. So this is what we do, and we do that for mid-cap companies in Germany. in different industries and this is what infinite mind is about Sebastian: Awesome. thanks a lot for the brief introduction, Stefan. really interesting. And I hope you didn't already give away the the big reveal at the end that technology is just 20% of what you need to do. But André Neubauer: Yeah. Sebastian: anyhow, I guess it's it's also a nice cliffhanger to the main topic that we're going to talk about today. But before, as usual, we jump to the main topic, we also like to discuss with our guests how they are working currently because everyone is utilizing AI, right? Some more for coding, some more for other general knowledge work tasks. And yeah, we'd like to do the same with you. So what's your tech stack? How are you working currently with AI? Stefan: Yeah, so basically what we what we do is we really try to eat our own dog food in in the way how we operate the company. So given that we we help companies adopt AI in their day-to-day work, this is pretty much how we also orchestrate and and and optimize our own work. So we try to be that AI native firm and we we look into How we can use the technology in the way we work, starting from research, sales, analysis, coding. So our internal admin work, obviously. So this is like the the our mantra that we try to adopt. And I think the biggest change that we see in the way how we work is pretty much that we now moved away from using like individual prompts or tools, but we really try to automate the process seamlessly with methodologies that are reusable. Right. So if we, for example, have like an approach to analyzing processes, that is something that we turn into a repeatable and reusable module. And this is pretty much our philosophy of how we approach that. Sebastian: Yeah, awesome. That that's really interesting. And so can you give a brief like run through how you would do that? So it it like you you figure out there there is a process that you can probably automate. Do you already have a standard approach on on how you do this or is this always a little bit depending on the on the process? Stefan: Well, this approach usually works when you already have a clear understanding of what the process should look like. And I think this is in general, we will talk about that later, important. So you need to have a clear idea of where you what you want to get out of it and how you wanna get there. And once you know that and you've done that repeatedly, maybe in the first time manually or like with less system support. So once you have that clarity, then it's much easier to turn that into whatever it is, into workflows, into individual skills. so that's usually the way how we do it. So once we have like a clear pattern and process and methodology, we turn that into a reusable pattern. Right. Sebastian: Yeah, makes sense. So I think we had one guest who had in their company a standing policy that everything that any employee of the company was doing for the third time, they should think about how they can automate this because then you start to see the patterns right and see the steps and then you have pretty much an understanding of what you need to do. Yeah. Sorry, didn't want to interrupt you. yeah, what what's your tech stack that that you're working with? Stefan: Well, so for for our customers, first of all, we're our philosophy is that we we are really very flexible because we we believe that we need to fit seamlessly into the t tech landscape and the tech stack of our clients. so but this is like for the client work internally. we we're pretty much now after we we experiment a lot, but we now narrowed down that we do a lot of work With Cloud. So that's our main, let's say, day-to-day working environment for everything from writing, analysis, document creation, all that. And this is this is basically the core. And around that, we we start connecting different tools. So obviously, because we l work with a lot of clients that are on Microsoft, we work with a Microsoft environment, so we connect that. and I think what is what's exciting is that Given that that the cost of building actually custom tools has gone so down so much, we are now also building our own custom tools, for example, in the CRM space, right? Because if years ago, if you would evaluate building a CRM on your own, well, I think that's a tough case. So now you spend like I don't know, a month or two, and you have a really solid piece of software. that is tailored to your process. So this is something that we are now doing more and more, of course, for the core elements of our business. So I think that's actually quite an exciting element of the tech stack. André Neubauer: also mentioned, well, like in the pre-talk, but also in, well, when we exchanged on the topic that Infinite Mind is AI native Stefan: Mm-hmm. André Neubauer: company. I think a lot of companies claim that you have maybe a characteristic. Why? So what distinguishes a classic company, so to say, from an AI native company? Stefan: Yeah. so I think first of all, just to make clear, what we try to avoid is really to be too black and white and like too buzzwordy. So I think being AI native I would say is a vision and an aspiration. So first of all, because I think it's also not very it doesn't make sense to do everything with AI. Right. So there is a lot of things where a simple automation can do the job, which is much more cheaper, André Neubauer: Hmm. Stefan: or you just don't want to do it. Right. So for example, if we communicate with clients, we believe that, of course, we use AI in the process, but we want to do that with a human touch. So I think being AI native is more of a vision. However, as I said, whenever we think and believe that There are certain things where we can repetitively do with AI. We really try to force ourselves to do that. And then I think there is another side of it that you also need to create the fundamentals that are needed to become that company. And one fundamental is obviously data. So we are a knowledge company, right? As a consulting company. So a lot of our IP is knowledge. So what we do. to get there is that we capture almost every piece of knowledge, no matter if it's said in a meeting, if it stays in the CRM, or if it's in a document. So capturing that and using that later on, I think that is one of the fundamentals that we believe is necessary to get there. Sebastian: Absolutely. André Neubauer: Don't want to already transition to the meat. That's why I would keep that. Maybe my follow up question for later. Sebastian: But but I think honestly, I mean, we th this is the most interesting part of this episode, right? And we can already transition into the meat, I guess. André Neubauer: Then may I ask my follow up questions, Sebastian, if that's fine with you? Sebastian: Yeah, please, please. André Neubauer: So I'm wondering, so my gut feeling is, and that's why I was really looking forward for our exchange, Stefan, is you put the nice headline, don't design AI for your company, redesign your company for AI. And I think it needs it so much. it's actually also, Also my gut feeling, when I'm looking at organizations, I see certain patterns and my belief or my hypothesis is that these characteristics or these patterns, they will change, they need to change. I think everything around processes, systems, teams. think everything will must look different. think must look different. You just made an, from my point of view, important point. You said you developed your CIM system on your own. I think half a year ago, time is running, but there was this ZAS apocalypse, right? Stefan: Mm-hmm. Mm-hmm. André Neubauer: where people said, okay, like now you can build everything on your own. And what I overall see is way less, a way less static environment. And this is true for systems. This is true for processes. This is true for teams. Unfortunately, I can't think of such an organization. That's why whenever I hear AI native company, I wish for someone is telling me, yep, Andre, this is how the future company should look like. Do you Stefan: Mm-hmm. André Neubauer: have some best practices on that? Stefan: I I first of all I chose that that headline, you know, you need to not build AI for your company, but you need your to build your company to to be an AI company, right? You need to build that for AI. And why is that so important? Because I think that there is this myth of of of this, you know, of this magical technology that will now come into your company and and and and transform everything and and create tremendous economic value, right? And I very much believe that this is true. However, I think that people underestimate what it takes to get there, right? And I think when when also when people talk about like how quickly change will happen and then what we do is we work With CEOs in of mid-cap companies, and we have these conversations over and over again. And I think once people understand what it takes to get there, and I think you mentioned a couple of things. So that you need to change the way you work. You cannot just take a human process and then put AI on top and think that this is now an AI native way of working. So you really need to change the process. So if you need if you change the process, all of a sudden, the people working in that process so far, they need to do something completely different, right? So you need to change the skill set. So we had just to give you an example. So we had we worked for for a business development team. So that team was building presentations for sales. So we automated building presentations, and what that team had to do. is they needed to orchestrate the knowledge and the data underneath. So totally different skill set. And and and and then you come to the point that you need to go through this change. So I think I still believe that that we will get there. I think that people don't realize how much hard work it is to actually get there. André Neubauer: Yeah, my observation basically is, Companies approaching AI as a technology, introducing us as a technology, and everything beside it stays. And then you have two very different setups which do not really work. great together, like something which is very dynamic, AI, which can be very dynamic, and something which is very static, processes and teams. And I absolutely get your view or your point on you need to rethink processes. What's hard for me to grab is the idea of how to build organizations because Well, like you could say, like the composition of a team is not different for sure. But as humans, think we require some, some, some, some staticness And this gives me really a hard time because yes, technology can be dynamic. Yes, processes can be dynamic. But what about the organization? Stefan: Yeah, so I think that's a that's a that's spot on and and and I think that the and I I will make maybe a step back and then I'll I'll answer your question. I think what we see or what worries me the most right now is that People are not catching up with the speed of change, right? And I think this André Neubauer: Absolutely. Stefan: is a fundamental issue that I'm thinking a lot about. And I think it so when we look at at past transformations, like they took like a hundred years, then fifty years, then maybe a generation, right? So using a computer for example. And now we have that acceleration within very few years. And if we look at a technology now, even like it takes a quarter and it's like a step change. And let me let me share like one observation. I was yesterday I I came back from Berlin. I visited the MyWay conference that's like a family owned business conference that was great by the way from from the pioneer in in in Berlin. And I talked to so many mid cap companies and people are just overwhelmed, right? So they don't know where to start. And everybody is telling you you need to adopt AI, you need to make that transformation, and that's the future. And if you don't do it, you will you will you will be left behind. And at the same time, these people they they're so they don't work at in technology. So this kind of change that needs to happen happens at a speed that is hard to absorb by by the people. Sebastian: Yes, absolutely. And and we had this in our show also several times, right? So that there are several things at play here. So number one, the the future is always unevenly distributed, right? So it's there are people who are adopting this at breakneck speed. And and even in in our domain, this is very different, even in in digital companies. There are differences or there are worlds between how I don't know, probably a more traditional bank would approach this that's fairly digital, right? But versus a very young startup that that is developing agentic software. I don't know. and yeah on the other hand there there are b businesses that that have a higher pressure, I think, to transform. So I'm I mean AI will transform everything. So that that is probably already visible and and most companies need to exposed to that, right? But then it's still different if you are a digital company versus if you are a restaurant, for example. Right. So there's a Stefan: Mm-hmm. Sebastian: different like AI will have a different impact on your supply or value chain, if you wish. right. And that that's I think we always need to keep in mind. And there there are people that say things like digitization already didn't work for big companies, right? So the transformation to a digital company already didn't work for for certain big corporates. and yes, I I do understand this perspective, even though I think to some extent most corporates probably have digitized by now. But then the AI it exactly, exactly. Either Stefan: I think those who didn't are not around anymore. As simple as that, right? Sebastian: they have or they are not around anymore. Exactly. And then on the other hand the AI transformation is much much faster or needs to go much much faster probably right at as far as we know currently and it's again right the companies that manage this will succeed and those that that don't will probably not succeed. and it's I think yeah that there are these these two perspectives right so on the one hand you need to I think you need to take this transformation head on and try to find a way the right processes also the right speed for the people you have in the org for for the culture you have in the organization, right? and on the other hand, yes, some transform sorry, some companies won't make it. And then new companies are popping up and that's already happening, right? And that will replace the old ones. that's at least my perspective. I we we've been moving a little bit into philosophical territory by now and we can get back to your experience, Stefan. from your consulting work so far, right? André Neubauer: Yep. Sebastian: Because you are working with these companies. I just wanna say to to Andre, I guess the the bad thing is there is no clear template yet that you can just put on companies. And I would also say that even within companies, like the engineering team, will have l different constraints than the HR team or the finance team or the marketing team. And all are like they need to find their own speed and their own tools and processes to some extent. That that's at least how I would summarize it for me. André Neubauer: Yeah, yeah, absolutely. Maybe just to let's sum it up, there is this nice, there's this nice pattern saying you can change an organization on process level, an organization level, on culture level. And I think this pretty much explains what we just discussed. Pretty straightforward. Like you can transform an organization with AI on process level. I think that is what companies are doing, right? Or what companies try to do. I think best practice here is not just copying the old process, but like, think it, rethink it. Like second is, and this is basically my question so far is like, how will the AI native organization will look like? Right? Like, so probably Stefan: Mm-hmm. André Neubauer: less people because like more of the, like say redundant stuff is handed by AI. And then, and I think this is probably the most important thing. maybe you need to start with that as the culture layer. You need to adapt Stefan: Yes. André Neubauer: your culture towards AI because everything will be super dynamic. And with that, I'm stopping here because yes, it's philosophical. I'm really looking forward for Sebastian to better understand what you have seen in your consulting work. So thanks for the discussion, Stefan: Well, what we see with companies that are really successful at going through this change is that it's actually a very simple thing, but it's it's so important. They have a clear narrative of where they want to go with their technology, why they want to do it. And this is a narrative that needs to come from the top. And this is something that has to come from the CEO. or from from the owner of the company, and that is setting the direction of where things are going and why things are happening. And it has to happen internally. And even more so, it also has to happen externally because you want to be in control of that narrative of what your company is doing about AI and about the technology. So I think that's the starting point and that is creating direction. And alignment because if you don't do that, people will fill that void and employees will will fill that void with fear. And investors will fill that void with interpretations that you might not know what you're doing or might interpret it in a different way. So I think this is the starting point on that path to becoming that company. Sebastian: Yeah, and maybe quick question on on this one, because I I wholeheartedly agree and I also have seen this play out, really. The the question is just like how do you or have you seen best practices on how to do this in a way that addresses the fears of the employees in a like good way and yeah, pretty much d drives the energies in in a positive direction or or le pretty much focuses the people on on the positive direction. Stefan: Yeah. Yeah, that's a good point because to be honest, I think they're good and b good and bad examples. So when I think at those leaders who did that successfully, I think what they do is first of all, they are authentic. and that also means that they clearly say what they know and they also clearly articulate what they don't know. Right? So I because otherwise you're you so you're moving in a field that is highly uncertain. So pretending you know all the answers, I think no one will believe you, and then then you lose credibility. the second thing is as I said. You need to create that that direction. So I think you need to be concrete about what you know. So you need you need to clearly articulate why this company is doing that. Right? Is it because you need to reposition or you want to to maybe change your your business model, or is it because you have like different considerations? So being clear of the why and the reason why I think is important before you talk about. use cases and you talk about you know what what we like to do in Germany about compliance and limitations. So why are you even engaging in this, right? And I think then you can talk about what you're doing. So so I think think this is the pattern that I that I see in in those leaders who are successful at at conveying this this this narrative. Sebastian: Yes. I wholeheartedly agree. And I would also say that the vast majority of people, no, probably all the people that we had in this show have been communicating in this way. So pretty much laying out, look, this is the future. I'm pretty sure that we need to adopt it, right? For whatever reason. But one reason is that our competition will adopt it. If we don't do it, then we are falling behind. Or We might have an edge if we adopt it now and we are faster than our competition, right? Or we can do way, way more suddenly than we could before just by automating road tasks that yeah, we just don't need to do any longer and AI can do right. So many, many reasons. And at the same time being very transparent about don't know the exact path pretty much. So how we get there and what exactly we'll do, it's just important to start and to yeah, get get a good feel for the technology. Stefan: Yeah, and I can tell you, I mean, as a leader, don't expect that everybody is is is is standing up and and and and and applauding, right? so I think that that will not happen. But I think this is exactly where where where you need to set the direction and you need to work on that, right? And you need to make people understand and take them on onto this journey. And that's a very hard thing, right? So Sebastian: Yeah, absolutely. Stefan: It takes takes a lot of leadership to do that successfully. Sebastian: Absolutely. And now the the the final point on this one, right? So I I do see for for tech leaders, I I see like how they can do this and how they can understand the technology well enough, but it's I I feel it's Stefan: Mm-hmm. Sebastian: or it could be harder for not so technical people, right? On the other hand, mm yeah, most CEOs will like have their networks, will have their news sources and will get some insights and will know how others are approaching this, right? So some to some extent they will also have knowledge on yeah, what this technology will enable and whatnot. Stefan: Yeah, I think you have no alternative as a leader Sebastian: Yes. Stefan: in our times to not deal with the technology and understand what it means or it could mean and how you want to deal with it and what the position of the company is towards that technology. I think there is no alternative to that. You cannot delegate it, you cannot give that to your CTO. You can work closely with your CTO on that, but you need to own that decision and that direction and that narrative from the top. Sebastian: Check mark. Yeah. So beforehand you mentioned a couple of pitfalls to us in the in the pre-discussion, right? So that was the first one. what's what's the second pitfall that you see? Stefan: well, I think the next one that that usually we see is that there are leaders and they're saying, Look, guys, you now need to use AI. And by the way, we are now rolling out this and that tool, very popular, rolling out co-pilots. and they start basically the whole journey. Upside down. So they start with a tool, right? And which I think is tempting because everybody is seeing this these videos and these promises and these use cases, how you know it automates your entire sales funnel or whatever it is. But I think the pitfall is that you start with a tool and you don't think thoroughly enough about what problem are we actually solving. And this comes back to the point that I mentioned in the beginning. You need this this notion of AI native. You don't do that for the sake of it or because it's fun. You do that because it creates economic value for your company. So if you're not clear about the problem that you're solving and you jump into the tool, I think that's like a very that's happening very often. And I think at the same time it's one of the biggest pitfalls because it derails the conversation. into a tool conversation and not into a change conversation. André: Maybe this is a misunderstanding, but I I need to challenge that. Stefan: Mm-hmm. André: because I think also earlier we said you need to make your hands dirty, right? The entire organization may need to make your needs to make your hands her hands dirty. Stefan: Right. André: so how how do you do this without having access to that technology? But but maybe I got also the point wrong in you saying yes, the capability needs to be there, like accessibility, like access to one of the leading frontier models for everyone, but it should not be the discussion on the specific tool. But yeah. So Stefan: Yeah. I understand your question. So I'm not saying that it is wrong to give people the tools. I you're exactly right. André: Thanks for clarification. Stefan: So I think it's very important to give give your entire organization the tool, and we will get to another pitfall when doing so. And and you probably shouldn't give them just one, but more to experiment. I think the point is that if you just give them the tool and you don't think about how the technology which problem the technology is solving for your company and where you want to focus your efforts in adopting that technology. That for me is is that that pitfall that I see that you don't think about what business objective. So just to give you an example, so do what is what if my strategy is that I want to go to expand into new markets. So then I should think about how can technology help me to move into a next market. So how can I make a market research and automate that? How can I create new leads? How can I contact these leads, for example, right? Making this up. But having this clarity is important because otherwise you there is no focus. André: Yeah, absolutely. Sebastian: Yeah. And I I think we heard a very good example of this from Francie from Mirgo when we talked to her. They had a an AI enablement session in the beginning. Once the I think it was the CEO who said, Okay, now we need to transform the company towards an AI company that has been after engineering. Obviously, had been tinkering with a genetic engineering for a while, but at some point they said, Okay, the whole company needs to become AI driven. And then what they did was they had these forward deployed engineers or AI enablers that had been very experienced already with the use of AI and had the people sit down, tell them a process they thought that could be automated, right? And and set together with them in order to set up the system and together with them automate this specific process, right? And that that also is very similar to I think Steve Yege wrote in in one of his articles about how Netflix is doing it. They're also doing it in a way that the full team, together with the manager, is undergoing a five-hour training with a forward-deployed engineer or several trainers. And they are actually taking specific processes from the daily work of the team and are looking at how to automate this. And this is really giving the the concrete Like guidance on how to use it in the day day to day work as opposed to only yeah, here you have a tool and now you can start tinkering and have fun with it, right? Stefan: Yeah, I think and and that's actually another another point that you're saying, and that's equally important, right? So one is being clear of what problems to solve, and the other one is how do I actually make sure that people are enabled to work with that technology, right? And that's a a different a different point, which is equally important because people are so I think the worst case is you just get the tool and then then You know, figure it out. And you would be surprised how often this happens, right? So everybody is getting this this this co-pilot license, go figure it out yourself, right? So that's I think the worst case. So it's like a gym membership. So you get the membership, but you never get go there, right? So the there there is no point. So I think this kind of adoption to to to manage the adoption and to support your employees. Of understanding the technology, giving them, for example, forward deployed engineers to work together. And by the way, this is also a way how we work because we believe you need to bring these cross-functional teams together, right? And to lower the barrier. And what we we always say is that your organization needs or your employees, they need to develop an intuition of what AI can do in their process. They don't need to necessarily be able to build that agent or to build that workflow. But they need to understand that and to to build that intuition to say, okay, this is what to what what AI can do for me and and and this is what I need. And then you can give them somebody to work with to build that solution. But that needs education, that needs training, that needs getting your hands dirty, that needs experimentation. Yeah. That's another very André: absolutely. Stefan: common pitfall that we see. André: Yeah, staying with your great comparison with the gym membership. I think I I think you can see two extremes. Like one extreme is not not going there, right? Not going to the extr Stefan: Mm-hmm. André: gym, but also training the wrong way. I have also seen that. I think it bit relates to the point Sebastian just made, like throwing the gym membership. But then like n having no trainer showing you how to train, right? you should not underestimate the the the interest and the willingness of an organization like or of people, employees to change processes or like automate processes. The point is like these are non-technical or non-AI people building stuff like in a way which is Usually not how you say that? Sustainable. I think that it that translates or that summarizes pretty well. And then you end up with like AI technology or like AI in the company, but done in the wrong way. You get what I'm what I what I what I'm saying or trying to say? I think one aspect Stefan: Yeah. André: is people not going there or not using it. And I think that's also not cool. What I think is even worse is like people using it without having understand. Stefan: You do the bench press wrong, you know? You do the bench press wrong without the coach, right? Yeah. André: Yes, absolutely, absolutely. Absolutely. Stefan: And and I can give you a very, very specific example where we that was very early on, where we literally had this experience. So we we worked with a client and we started working with a team early on, like on on, you know, setting the business goals that we want to accomplish and then narrowing it down to the use cases. And we didn't make good progress. So The input from the client was not very good. And we were really wondering what the problem is. So we decided to stop that path together with the client and invest in like a very focused upskilling program. So what we did is we did functional trainings with the different areas with the finance team, with the product and dev team, with the editorial team that was a media company. So very specific. Showing them what the technology can do for their use cases. Right? And guess what happened afterwards? So you you immediately, even after the sessions, had people coming up to say, okay, now I understood I can you know collect this data and transform this data that way. Does it mean that I actually can do this and this and this? And then you had like they understood, they did develop this intuition even after like a like a very con like condensed training of what the technology can do and then they started thinking, right? And and and this is this is exactly what you are describing. So you give them the training, you show them how, and they develop this intuition and then they can come up with ideas. Sebastian: Awesome. And and how did you do this training? So did you take examples from their day to day work or or just Yeah. Yeah. Yeah. Stefan: Yeah, super super hands-on and ta tailored. So it wasn't like like this generic training. It was like, okay, so you are a designer. So let's look into what are the state-of-the-art design tools, what you can do with them, and how you can use them, how you can create illustrations, animations, videos, etc. etc. So it was super tailored, and that would be also my recommendation about upskilling. You need to do that in their given work context. with the use cases that they work on a daily basis. I know it's it's it's gonna be more it's it it it takes more effort, but it pays off because it has to be tailored to that that realm and that that that area that people are working in. Sebastian: Yes, yes. All right. cool. So we're already quite advanced in in time in our episode now. I think you still have some more pitfalls on your list, right? Let's try to manage to get all of the pitfalls in for the main section before we then go over to the last two sections, the the reality check and also the predictions one. Stefan: Sure. Yeah, let let me pick I'll I'll pick two just to be brief. I think and they're that they're they are very important in my view. So we talked about processes in the beginning, and I think one very common pitfall is that people take the process that exists and then they try to automate that, right? So let's take a process, people are doing certain steps, and now okay, great. We can take that process step, for example, I don't know, doing the analysis of of of data and then we we automate that, right? And The first step is that, first of all, if your bad process or if your current process is bad, it will become worse for sure. And the second thing is you need to really rethink the entire process with that technology. So don't try to just automate the existing one, right? Because I think there's like a difference between an AI enhanced process and really a a process that you design with AI enhanced. as as a first principle. So I can give you an example. So in in in a media company, right? So you if you create, for example, a video. So in the past you had like an editor writing a story, then you had like a video editor creating a video, and then you had the social media team posting that on social media, right? So if you now start automating each step the way it was, you will automate three teams. Versus if you redesign that, you will maybe have an editor that is creating the story and then everything in that follows will be automated and you actually don't need these hands off to other teams, right? Or you involve them in a completely different manner. So that's what I need. That that's what I mean. So I think don't Don't try to optimize the status quo. Try to rethink based on what you want to get out of it. So I think this is Sebastian: Like Stefan: one. Sebastian: yeah, two two thoughts on that very very quickly. One is when I think in the area of digit sorry, the yeah, the the era of digitization, we always said if you digitize a shitty process, then you have a shitty digitized process, right? So same thing. And also, Stefan: Still true. Sebastian: yeah, thinking thinking similarly, if you digitize a process where you I don't know, file documents and you now say, okay, we scan every document and we put it in a in a digital folder somewhere. as a scanned PDF, then you also you don't have the the full value of having the digital document that you can then index and you can make it searchable and everything, right? So I think it's it's a very similar mental model that we can apply to to the current transformation. Stefan: Yeah. André: Absolutely. Stefan: Yeah. And maybe the second point that I would I would pick here is what we see as a is a high risk is that companies so there's this number always going around that like whatever, ninety percent, ninety two percent, ninety-five percent of AI pilots actually don't have that economic value, right? And I think the point is, or what we see, is that one of the reasons for that is that many companies are still stuck in that pilot phase, right? And I think we all know it's nowadays it's so simple to build a compelling showcase and pilot case. And I think what organizations need to have a plan is how do I scale that into an organization of 2,000, 5,000 people. So what do I need to have in place in terms of governance, right? In terms of infrastructure, in terms of permissions, right? So and and the list goes on, right? So I think having thinking ahead and knowing what it takes to scale and putting that in place, also like a governance on deciding, right? Or and I which is becoming more and more a governance on cost, right? So I think there are so many aspects to scale versus just to build something that looks fancy. and I think this is something that companies are still figuring out and that is blocking many companies from achieving that value. Sebastian: Alright, so your your your point is that this needs to be thought through from the beginning, basically, that already when you're running the prototype, you need to think ahead of all the pitfalls that come with something that needs to run at scale. It's pretty much like building your own little CRM for yourself, right? Which is perfectly fine if you need to, I don't Stefan: Mm-hmm. Sebastian: know, fix a bug here or there, or it's not working 100% all the time. But if you wanna roll this out to thousands of people, yeah, then it just needs to be reliable and secure and yeah, available all the time, right? So yeah. Stefan: Correct, yeah. And and and as I said, I think it's not just a technology prerequisites. I think it has a lot to do with compliance, with with you know, governance, like I mentioned, permission rights, like what kind of data may I access or not. So it's I think it's it's not only technology. I think technology is an important piece, but I think being aware of what it takes to scale, and also like planning that for that, that the this needs to happen, right? Also in your business cases. That that's my point here. André: Yeah. Sebastian: Yeah, André: yeah, and my takeaway from that is actually like you next to that you need to have a plan, right? Like so what is the step plan? You you you introduce the tooling, you give room for learning, then you start with some prototypes, and then you bring this like you you professionalize so to say the company, it takes time. My my one of my Biggest learnings is it takes time. It doesn't happen in the afternoon between three and four. It takes time. I think it's one of the so if you would ask me what is the what what is my pitfall number, maybe not number one, but within the top three is like people think that's an easy win. You just need to approve the tooling, then you need to have some training, and then like it will it will automatically transform no way. Yeah, Stefan: It will go through the roof. Yeah. André: yeah. No way. No way. Yeah. Sebastian: Absolutely. Stefan: Exactly. Sebastian: Yeah. Stefan: It's I mean that's the the whole point. I think we're talking about a con a change of a company across multiple layers from culture over processes over organization over technology and data, obviously. And this is a complex thing, right? So and I think that's that's what what companies need to realize when they embark on that journey, because otherwise they will just get like a very shallow benefit. And then you have these discussions of is that technology actually adding value? Yes, it is, but it takes an effort to get there. Sebastian: Absolutely. All right. I think that's a great closing line for the main section right now to the André: Absolutely. Sebastian: two smaller regular segments of our show. The one the first one is the reality check. So usually we ask our guests for what the fucks and and what they think is hype or are hype topics currently. So do you have some what the fucks and hype topics for us? Stefan: yeah, I think there are there there are a lot of there are a lot of hype hype topics. So first of all, What I always wonder is that frontier line of what what a model can do and what it can't, right? So that there are like so intuitive things that the model still cannot accomplish due to the way how they are built, right? Because we we know all the kind of spelling things or viral things where There's actually one funny guy on on TikTok that is always pushing the models and and you know handing them over and pretending that they're talking to his boss and then he's calling in sick and things like that, and that that it's totally awkward, right? In voice mode. So and I think what but I what what I actually really is still you know surprising me is that we still cannot reliably say if a model is going to accomplish a certain task perfectly. Or it will just completely fail, right? So I think this is still something that is that is for me always something that's that's on my mind, right? That's that that fascinates me every every time, right? Sebastian: That's that's very true. Yeah. I I mean there are certain patterns that that we can probably make out already, right? But then I also feel like sometimes I have a very small request, something I want to get done quickly, and then I'm ending up like I don't know, an hour trying to fix the the issue and get it to do what I want. Absolutely. Yeah, can relate to that. Stefan: And and maybe just on the hype topics, because I just need to to to tackle this, is this current conversation about you know the risks of AI and humanity being destroyed? I mean, just by the technology. I mean if you let's think about that way, if the companies and their CEOs would be personally liable for every damage. That the models do. I wonder if we would have this conversation or if they would find a way that these things are much better under control as they claim to be, right? Because in the end, that's still that's technology being built by people. And I'm pretty sure that people can put controls on that technology if they wanted to, right? And I know there are more conversations about is that is that slowing down innovation, etc. But I think purely from the point of is that technology going to kill us all? I think if you think that way, I think there is certainly a way to govern these systems in a way that's the conversation will be a different one. Sebastian: Yeah, for sure. The issue with that though is that the current the current I think regulation, at least in the US, is that only humans can be held liable of their actions, right? So if there is a security breach, then I think the the rule is something like the The a human actor must intentionally or knowingly access a protected computer without authorization. Stefan: Mm-hmm. Sebastian: and this yeah, is basically if you have yes, exactly. And and the thing is Stefan: It's a regulatory topic, yes. That's what I mean. Sebastian: and and the thing is that exactly, yeah. And the thing is that in the current environment where you have this genius on top of the US government who who just probably did doesn't understand what's happening there or doesn't want to understand, I don't know. he will not or this will not be changed. I I think the the the the perception there is currently they are in an arms race with China and hence they must not stop the this this race, right? And I think it's it's really hard to make a case for a change of the regulation without something really bad happening. hence I think that the law sorry, the the frontier labs themselves taking action is probably the better chance of of of not something severe happening before any regulation will change will be changed, right? Stefan: Yeah. Yeah, you're absolutely right. And I think the point that I wanted to make is is that it's it's not something that we that is, you know, happening uncontrollably. I think this is something that can be regulated if we wanted to, right? So and Sebastian: Yeah, yes, yes, yes, I agree. Stefan: I and I think it's also not helping the progress because if people are scared scared, if the public is scared and you race and you don't you stand up and you don't want to have the data center like in the US in your town. I'm not sure if this is not stalling progress as well. So I think that's that's an interesting discussion Sebastian: Fully Yeah. Stefan: to have an eye on it. Sebastian: Yes, fully agree. And in in our last episode we actually discussed the AI apocalypse topic. And there we also made the point that we don't think that the what's it called? The People's Chinese Party or whatever, right? They will probably not lose control André: C C P Sebastian: or s the C C P exactly. They will probably not allow an AI that is not aligned with their with their wishes, right? So they will find a way to regulate this. And yes, the US would certainly also be able to regulate this. Absolutely. Fully agree. All right. thanks a lot. Yeah. So that was super cool. And now final topic, prediction. Do you have a prediction for our listeners? Stefan: Prediction. Yes. So as I said, I've I've spent the last two days in Berlin talking to a lot of family run businesses in Germany who are industrial companies, right? And they now start implementing they they now start implementing AI more and more. what I see coming because There are signs, obviously in the US, but now more in Europe, is that I would expect that within the next 12 months the question of cost will become like a very central one once companies start to scale that technology across organizations. Right. And we see that with the tech companies limiting token spend. And I think I I I see that discussion coming once some crazy bills start piling up. So we as a company at Infinite Mind, we are actually preparing for that with an with an offering towards our clients of how to actually manage that cost. how to be more efficient when you look at how you build systems, how you cage, how what models you use, what infrastructure you use, et cetera, et cetera. But I think that will be a hot topic in the next 12 months, where all of a sudden there will be a report in the media of companies spending a lot of money and being becoming more cost conscious. Sebastian: Yes, fully agree. I think we as you mentioned, right, we saw this in engineering already, where Uber and Spotify now say at least that they've reduced their token spend by ninety percent, at least on a per engineer and month basis. I think that was the the the message there. And we'll see this in other disciplines as well, for sure. Alright, thanks André: Awesome. Sebastian: a lot. That was a really cool discussion. yeah, thanks. Stefan: Thank you. I enjoyed it a lot. Sebastian: Thank you. André: Thanks, Stefan. Sebastian: Bye bye. Stefan: Bye.