Latest / Beyond Vibe Coding / #S2.06 Your AI Transition Does Not Stop with Adopting the Technology - Stefan Schubert-Peters
Transcript
- Stefan: most of our engineers are already using AI some more some less right, but but everybody's using AI So you thought yeah, well, they are already at least experimenting if not fully integrated into the flows, but also experienced engineers Seem to have had like a down moment in that situation and what I then realized Is that's actually two changes happening at the same time André Neubauer: Welcome to another episode of our new season of Beyond Wipe Coding, Partnering with Impala Search, the go-to tech and executive search agency in Germany. In this podcast, we explore the transformational change in software engineering and acknowledge work in general. I'm Andre, CTPO at Trusted Jobs. Great to have you back. This time, unfortunately, it's just me. While my co-host, Sebastian, is enjoying a well-deserved break, I'm talking to Stefan Schubert-Peters again. And we are covering a soft topic today, relevant, well, I would say for all companies going through the AI change. It's basically your AI transition does not stop with adopting the technology. So without further ado, let's jump into the exchange with Stefan. Enjoy! Hey, Stefan. Welcome again. Great to, yeah, speak to you again after we talked, I don't know, a month. No, it's not a month. More two months, three months ago. Back then, actually in German. So I think it's just fair that you introduce yourself again so that everyone knows who you are, what you're doing. And then, yeah. Stefan: Hi. Yeah, two, three months. André Neubauer: We kick this off. Stefan: I'm really happy to be back. Hi, Andre. Feels a bit like I'm stepping in for Sebastian who can't be there today, but totally good. I also have a... I just real-time looked it up. We met the last time in February, so this is already five months ago. No, it's all good. I was also thinking two to three months, but actually it was February. Good to be back. André Neubauer: shit. Shit. Time is running. Time is running. So, what's your role? What are you doing for a Stefan: So, yeah, just to recap, right, I'm CPT at Falmo. We're a big mobile veterinarian startup in the whole of Germany with over 100 vets. And yeah, I'm leading the product tech team, ensuring that we can scale because we scale with tech. And yeah, I'm leading that small but fine team. André Neubauer: To those of you who are interested in what Stefan's tech stack is about, think you can listen to the episode from February, since it's like five months ago and back then it was also German, so maybe not everyone can understand that. Is there maybe an update for us, Stefan? What has changed? Stefan: I actually think a lot has changed. When I'm when I'm correct, there was was February. So in December, I think Opus four, five or four, six came out. And back then I switched to Claude, like most of the people, I believe. And, and in the meantime, we've all had our ups and downs with Faber, right. And now I'm actually mostly working on on on codecs. So I've completely switched over to codecs in the time when failure wasn't available. And the reason I switched is because codecs was more predictable. So that's right now my, my stack of choice, which is also interesting because I think the last time we talked about if it's easy to switch between environments, and we've, we've made it switchable. And that actually works super well. And Yeah, so and I'm super happy with the GPTs, especially now with GPT 5.6 but also 5.5 was way more strict than Oppos was and that was one of the reasons why I then stuck with it actually. André Neubauer: Okay, today we want to actually cover another topic, maybe just to maybe you can spoil a bit because you highlighted it so much. You switched from Claw to Codex and you made that work. So is there a secret sauce, something you want to share in a nutshell, maybe for listeners? Stefan: So what I what I did right from the get go is is to always make sure that the that the the cloud MD doesn't contain the stuff. But basically, the cloud MD always has just as a two liner saying, all of the stuff is in the agents MD. So if you want to do anything, do it there. And that from the get-go also then worked with codecs. So I always tried when the limits were gone, I always tried codecs. So I was pretty sure that it was working. But when I completely switched, I also had to make sure that the skills were working. And one thing that I started with one or two months ago was actually using hooks a lot because like skills are more, I would say, are not really a contract, but skills are more something that lives in the context. And if the model chooses to forget, What's in the context? It also is forgetting the skills. By the way, codecs also better at remembering the skills. But hooks are actually an enforcement. So what was totally helping is making sure those hooks are also working in codecs. And it's pretty simple, because all you have to do is register them in the codecs environment. And then they are just the same technology, essentially, though. So they are scripts, listening to certain tool calls, like for instance, before pushing to GitHub, do a security check to an architecture like based on those conditions that you do. And I've also recently started experimenting with expectations. So basically taking the gist out of the... out of those hooks, like for instance when one hook is making sure that my change has between 85 and 100 % code coverage. And this is an expectation that I'm automatically generating and putting back to the agents MD again to make sure that Codex is always aware what it's checked against and so I have less agent, like basically agent loops. Because then it would come with the code, coverage is only 75 % and so like rightfully the coverage is too low so I now need to make sure the coverage is higher right and if you formulate the expectations right from the get-go that works and I've also started and that's super interesting in injecting those expectations with a hook after the compaction because you can with hooks or with other configurations and Claude and in Codex actually inject things into the context again. And so I started experimenting with that and that works pretty well. So I'm super happy with that loop. It's efficient because of that. And yeah, I don't look back and like right now switching seamlessly works really well. André Neubauer: I think two good learnings, right? First of all, helps you to not mess up context. Second, it's probably reliable. So deterministic. was just checking because well, I've, I've, I've heard it for sure. I, it's, it's, it's not the first time, but I've never checked whether this is also available with Claude and seems like it's not right. It's, is it something very specific to code codecs or Stefan: Yeah. You mean hooks? No, no, no, no hooks. No, learned about hooks in the reason I learned about hooks is because I was researching skills and skill distribution mechanisms include. And then I discovered plugins, but plugins also contain commands and hooks. André Neubauer: Am I just? Yeah. Yeah, yeah. Yeah, I'll get you. Stefan: And so I tested hooks and it's pretty amazing because you can hook into any, like you can like listen to when a Go file is edited. Before you do that, do this. And so that's basically how you can make sure your agent does something without having a skill loaded upfront because you might not know if you need it. So it's like that's really effective actually. But it's slower because it always needs to spin up those agents first. André Neubauer: Sure, sure. But I think since, well, you can work in parallel on certain stuff, it just increases the loop time, so to say. It's a larger loop. Well, if you're used to work in parallel, I think that should not be that much of an issue. Stefan: No. Yeah, and you... Exactly and you're getting like better one-shot results in the end of the day because like certain quality qualitative criteria met from the get-go and it's way easier than to To like you don't need to iterate as much anymore in the end because because André Neubauer: Now I also found it in a documentation. Shoot. Stefan: Because what you did before, I always have to remember that why it's really good. Before, what you did is you actually discovered that the coverage is not good enough because some PR commit hook, or some PR hook in GitHub discovered that. And then you got back, copy pasted that feedback to Claude, right? And so your loop was even longer because of that. And now it's all happening on my machine. André Neubauer: Yeah. Cool stuff, cool stuff. So, but today we want to talk about more of a soft topic, so to say. So you brought a story. You want to share it or shall I introduce it somehow? Stefan: No, no, it's all good. It's actually pretty, to be honest, because I like individual contribution so much. So I code a lot myself. But this is a management topic. So it's like really not my home turf. Maybe it's even maybe it's maybe it's more yours, Andre, I don't know. But what I André Neubauer: Let's see. Stefan: I actually stumbled over something in the last three weeks. One of the teams in my company had a retro. And then I got feedback from our tech lead and the feedback was that it was a pretty down, even toxic retrospective. where the mood was so low, everybody complaining about AI and where it brings them and about job security and stuff like that. And my initial reaction, my gut reaction was to be disappointed because I said, well, I tried to lead the topic for like over a year, you know? And that... But obviously as a manager, you can be satisfied with just being disappointed. It's a negative, unconstructive feeling. So I gave it a couple of nights and thought about it. And then actually an old management training topic came back, which I don't need that often in a small startup, but actually here it's super useful. And that is that I realized, okay, people are just going through change. And which is really surprising, right? André Neubauer: shit. Stefan: Change is obvious, but what does change mean? And so I researched again what this like change curve actually is, right? So where they initially they realize it they don't accept that something is happening and then suddenly they're like they they notice Okay, it's really happening and they get through this depression curve, right? Which is like the valley of tears until they finally accepted start experimenting with it and and it's one point of time integrate into the change, right? So that's like the the tip change curve that you learn about in some management one-on-one training a long time ago. I'm sure you learned it too. As I remember we had similar trainers. André Neubauer: True. And I can also add everyone is traveling this on its own speed. So the larger the organization, the more difficult it is to, well, ensure you're at the same, at the same state, right? You actually always need to keep in mind that some people are faster and some people are slower and like everything is fine, right? Like everything is okay. You just need to accept that you can. Well, I think you can try to accelerate that, but I think you cannot overcome this, right? You can say like, it doesn't exist. So it. Stefan: People have to go through the valley of tears. For some the valley might be lower, but it's because of the person, it's an intrinsic thing. So some might be faster, for some it's just a sleepless night. And for some it's really weeks of having to process it. But it didn't really explain it, so it stuck with me for a while. Because... André Neubauer: Yeah, absolutely, absolutely. Hmm. Stefan: Because actually most of our engineers are already using AI some more some less right, but but everybody's using AI So you thought yeah, well, they are already at least experimenting if not fully integrated into the flows, but also experienced engineers Seem to have had like a down moment in that situation and what I then realized Is that's actually two changes happening at the same time which I wasn't aware of and it's super important to be aware of that because otherwise you cannot tell people and you're just cynical. And the one change is the obvious one and that is okay like you're now using AI for software development but once you do it and you start being successful and you read and you continue reading the news about AI something else happens and that's the realization of how your job will change and how your job will change in the future. And that was pretty revealing and it was super clear that probably most people in tech are right now. are like right now somewhere at the edge of that valley, right? So some might be already out and that's probably the leaders, that's probably people like you and me. Like I can remember I had this short valley of tears moment, I don't know, like a year ago when my CEO started talking about AI a lot and I needed to adapt my role and think about it a lot, right? But a lot of people, like a lot of engineers didn't start that early and so they are that right now. That was a very interesting revelation. And then I discussed it a bit with our managers in the company and I think the interesting insight there is, as you said, how can I help people get through that valley. And that basically means first accepting that they're in it and that it's just like a personal thing that they need to resolve. So you can basically help them understand where this change will lead them and encourage them that the company they're working in is here to stay for them, right? And why the company needs them. just like basic, like creating a safe environment, basically assuring them of the safe environment. And also, but still asserting that this is here to stay because otherwise they might ping pong back and forth in that valley. So like really pushing them through the change curve, helping them along, supporting them. So that's like what we did. And then the study came out, which was funny at the same time. so like a lot of people are listening right now to this Lenny's product. André Neubauer: which Stefan: podcast, right? And I do too. So listen to all these inspiring tech leaders from different companies, but also the AI companies. So if you're listening to that, you're very much ahead and you know what's happening. But actually, he and Noam Segal published a study, a large scale survey of tech workers in 26. And they do these studies, like on a regular cadence. And they discovered that right now, the tech industry is going through a basically 50%. of the respondents feel amplified by what's happening in AI and for them it's like exciting and everything. And then the other half is having second thoughts. They have fears about their role, they feel destabilized, they feel diminished even. And it's like mainly a burnout thing, but it's or like the fear of job insecurity. But it's almost close to burnout because the main worry that most people have is that they're just asked to work way more with less resources using AI and that this is kind of pushing them. Yeah, it's just a very exhausting feeling. So one half of the industry is super energized and amplified and I can relate with this crowd very much. And then the other half is on the brink of exhaustion or fearing that they might be exhausted, which is dangerous, André Neubauer: And I think it's also, it might be also a combination with identification, right? Because a lot of people are identity, well, not just in tech, but also in other disciplines, like people identify themselves with a certain job, like with certain activities, right? They are proud. And now to a larger degree, especially in software engineering, this work is, is, is, is gone, right? So still there, but like it's not the responsibility of an engineer anymore to write the code like someone else is doing this. So if you're into, don't know, clean code, architecture, all that stuff, right? It's not a hundred percent gone. But like it's not your main job. and accepting that change, I think that is quite a stretch for people. So I can totally relate to that. But maybe step back before we get back to Lenny's podcast. So first of all, or Lenny's newsletter in that case, first of all, I like really the separation because if you look, companies usually measure adoption, right? And they say like, now we have an adoption 80 % or 100 % of the people are using this and 90 % of the code is written by agents. So we are done, right? And I think that is a false hypothesis. So as a company, you say like, okay, we are through that change. Like it's all good. Like we can move on. But the reality is you're actually in the middle of that change. I really like when we prepared for that podcast, I really like your observation. I was really blown away because... Stefan: Yeah, exactly. André Neubauer: It was, it became so obvious to me. So for, for us at Trusted Jobs, are a bit larger organization. for us, change management is, management is a bit more important. So I'm aware of that, but not to that degree. So, um, like that is a, actually you could say that is a second separate cycle, right? So adoption is there, but now you need to face the truth and say, okay, like, what does this mean? Right. Because it's not just like, no, that's something some, someone else is writing the code, but. It's the processes, the responsibilities, the skills. This is changing. Yeah, so a lot of stuff. Sorry, but I didn't want to interrupt you. I just found it so important to give it a bit more visibility. Stefan: Yeah, good. And but I'm curious when you like because I don't work in a bigger company right now, like do you how do you see it across teams? Are there teams which are already pretty far on both change curves and some are back or is it like really on a super individual level? Like how do you perceive it in a bigger company? André Neubauer: very good question and I think there is not this single answer. I would tend to say it's It's more an individual thing rather than a team thing. So individual, like everyone is traveling through this journey on its own, but for sure, like given the team context, you may be a faster or a bit slower given like the input you get from the team. So there is definitely a dependency or a connect to that, but I would still say it's more the individual thing rather than the... the team aspect, actually also given the fact that people have different belief systems. So if you look at teams and you usually have the ones like who are more the builder and people are more really into code and software design and stuff like that. I think these two groups, to use these both stereotypes, they would travel with very different speed. And we know these people. from our past. So there are people who are able to adapt pretty fast and for others it might be an issue because well, like the stuff they're very good at is now, yeah, is not there anymore or not to that degree. So I would say it's an individual thing. Stefan: Yeah. And I also think it's probably, there's probably also like an inverse. conflict there like or a paradox because I think for most managers, so for most people in manager roles down to the engineering managers, AI is probably a big revelation because they suddenly feel so productive again, right? So I would say for most engineering managers or upwards up to the CTOs, they say suddenly I can do things I can have a direct impact again. And they can do it with like their partial attention that they can give like just like one day a week or something like that you can suddenly do a lot and it's a big disconnect to the individual contributors for whom that was the bread and butter job and now they see okay like my my my CTOs now also contributing PRs like apart from the quality which might be bad this is this is also weird right so somebody who had a professional distance before and was mainly giving a feedback like I think very involved CTOs might have looked at PRs and just reviewed code and stuff like that, or maybe had like a little tooling responsibility. But most of the time one was your domain and your manager had the managing domain, right? And now suddenly your manager that's giving you reviews, performance reviews is stepping on your toes all the time. It's super interesting. And then imagine you are the one being managed. You're just like you finished university in 2018. And you coded for eight years and this end like you just build up like you maybe you've been able to level up until senior engineer. You got your promotion and then one year later you are pushed into this change. What does it do to you? Right? Super weird. Like now thinking about it, but I'm thinking about this for the first time. I didn't like with because of this change thought, right? I didn't have this thought before. So that's what I mean. There's also a huge disconnected seems. Maybe at least for me. I don't know maybe other city also a bit more aware André Neubauer: Yeah. I would say this regarding the people with less experience, that's 2018, it's 80 as well. Like it's not that it's a bit, right? But it's not like the, how you say, not the veterans. It's like they were, I think from my point of view, from my observation, I like, there are pretty good cases also at Trusted Jobs where especially people with less professional experience, manage that change pretty good because they're used to it, but they're able to adapt to it because if you don't do certain stuff for a decade or so, you're still flexible enough also in your mind to try out new things. I would say the people in between, so not the veterans who maybe also have seen other larger shifts. I think they're also good, but the people in between, right? So the, maybe eight years, maybe 10, maybe 15. So they, I could imagine they struggle because for them it was all the time software engineer, right? And they're more or less very much used, or I'm not saying like they can't adapt. Still, there are different stereotypes. But I will see more issues in these groups rather than the juniors or the very, very seniors. Maybe we can get back to the survey from Lenny's newsletter. Well, I would not say surprised me, but maybe we can talk about it for a second. There seems to be no fear on the management level, top management, founders. For them, it looks like they already passed this valley. You noticed that as well? What's your take on that? Stefan: To be honest, I didn't really notice that, so I didn't pay attention to that, but it kind of makes sense, Like maybe for some... middle managers, it might have been a big question. But then they probably also just saw that it helps. Like, because typically, who has the context in the company, right? Like, who is able to connect the business and the tech side together? That's typically the middle managers, and they can now also feel productive. And they are on the front of the change, if they're not like in one like dark corner of the company, you know, but but if they like informed and aware of the news, especially the founders, right? So I think for them it must, it's like for founders, it's probably more like a gold mine than like a gold rush situation, than really big fear unless you do something that is super commodity and is now basically made irrelevant by AI. Then I must imagine that this is also going away. But you've probably seen the stats, right? So. like the job market in those kind of companies that are a little bit more modern is going up. Because it turns out that with AI you now need experienced engineers who can make it work and who can connect the systems together and everything. I think that's also the reason. And they've probably also gone... Like they've seen their fair share of irritating change in the past already. And they just think, yeah, it's just a technology change. yeah, and maybe like the layoff at Microsoft, I don't know, like, it's like, you've seen that probably, right? So the 4000 people is like, how much are like, how many, how many percent like 2 % or something like that of the workforce or two and a half percent. That's Yeah, it hurts for every single individual that that like position that they leave. Then that they need to let go. But yeah, so I think probably it's just that kind of thick skin that people have developed. And as I said, founders are probably relishing in the in the prospect of all the interesting new things they can play with. André Neubauer: regarding Microsoft or like the other large US tech companies, make them too far away to maybe have a solid opinion. But I have an opinion. And for me, it's also the fact that these kind of layoffs are more or less normal. So you do not have these kind of large, large, large layoffs in Europe. Well. Stefan: Yeah. André Neubauer: if you are in automotive and the really large companies, this also exists. But I think for the tech companies, it's more of a, from my point of view, something more of a pattern. So to stay lean. it happens every now and then. So it feels a hard word, a natural. Everyone who's affected by that is like, this is very sad, but I don't know, if I look back for the last years, there's always increase and there's always a layoff waves. It feels like you're alternating between, I would not say extremes, but you over stuff and you try to get back, right? You realize it's getting too slow, then you stuff again because there is a certain demand. Especially if you look now, everyone is, we're getting a bit off. of the track, everyone is now absolutely off topic, is deploying front-end engineers, forward deployed engineers. So it seems like engineers are very much needed again. Stefan: Off topic. Yeah. André Neubauer: Well, let's see. Let's see. But I found a point in the survey. It's point eight. Founders are still the happiest people in tech and small companies are still the best place to work. Both findings replicate from 2025 and both are statistically robust. The nine, the point nine, I think is maybe we can talk about that in a, for a second, you covered that earlier already. Managers are still the biggest. lever for happiness. Manager quality remains the strongest driver of burnout and one of the strongest driver for everything else. I think in the very beginning you mentioned that managers already passed the valley of, I don't know, uncertainty. To be very, very honest, at least for me, I would say I'm not through that valley. I do not have that much of a fear, but I have a lot of unclarity. And I think it's also fair to share that with the organization so that people do not feel like this is their issue, right? That's something in their role, this is an issue because they do not have certain... capabilities or such. think it's just normal that for people, as you said in the beginning, like everyone is going through that change. I could also imagine it's actually causing the wrong reaction if you, as a manager, just like behave as everything is normal, right? So it's not that much of a change. All good, let's move on. It changed at that scale. I think it's just fair to show vulnerability, uncertainty. Sure, you need to have a plan. But on the other side, showing that it's also touching you, I think it's just fair. Stefan: Yeah. But is it the same, like if you say you are not, like you're also not through the change and I would probably say I'm, like, I would say I'm probably at the end of the valley in terms of job, like in terms of these kinds of worries, right? So how is my role redefined? Also largely driven to the things we've been doing in the company like right now we have having an AI July which is basically a program which is like where we do 20 where we do 20 small AI projects with like for non Non-technical people. So basically like people in planning, which are planning how that's work, they need certain tools, right? So and we're taking like, we're sending an analyst there or an engineer, which is building some AI tools for them. So and these kinds of things that like for me, this is just okay, I think I know. what the role now is, right? So you basically take the old CPTO handbook and not throw it out the window, but you rewrite it with way faster cycle times because you can do more with less now, right? That's like my way of thinking about it. And that's also how I think that I'm still needed if you wanna talk about insecurity or, but I'm curious, like how do you think about this change? for yourself. André Neubauer: So I... I do not have that much of fear. this is, I think this is not what I'm talking about, but uncertainty what the future role will be because we are in the middle of the change. If you look back at the beginning of the talk, right? We talked five months ago and what has changed on the technology side, which then has an impact on processes, has an impact on Stefan: Yeah, okay, that makes sense. André Neubauer: organization which had an impact on responsibilities. think roles are still unclear. I can deal with that. But I can deal with that because I have been through, I cannot say similar change, like change. And I think just sharing that and also sharing the fact that you think about that. Stefan: Yeah. Yeah. Yeah. André Neubauer: with your organization, I think that hasn't can have an impact just showing that everything is good. Like we will get there. it's, it's unfair because then you are denying either you are dumb or you're denying that something has happened. and, and back in the days, we call that podcast, technology is a people business. And I very much believe so. again, this, this is a massive change and you need to. Stefan: Yeah. Yeah, fair point. André Neubauer: which actually brings me back to, I'm bad in remembering episodes, but we talked with Franzi or Franziska from Miro some episodes ago. And I think she also mentioned that, I think it was the CFO who shared his personal story, like, so his personal change story, so to say. I think that could be a good pattern. And I also have another pattern in mind, but maybe you want to jump in here. Stefan: Yeah. No, I get what you're saying. So when you say you are not quite through, you probably mean, I'm probably through the change curve. So I have accepted that a lot of this happens, and I have also accepted that there's this insecurity. But this change is here to stay, right? So sorry, what I mean is, it's not that the change has happened, and now it's passe, right? But the change... will happen in the next couple of years. this like uncertainty is just due to the fact that we don't know how far those LLMs will bring us and when the LLM wave is over and maybe the capabilities are flattening out then there is still the Jan LeCun world model movement, right? So which like which is kind of the next promising step but it might also be always away 10 years like nuclear fusion right so you don't know if that's gonna come but for sure but for sure like we will we have at least another five to ten years of change ahead of us that is continuing to erode our like our like minds of people is continuing to nag people about stuff. My point is more like, think we now feel somehow comfortable in this new world that is changing now faster, kind of like the mobile first movement back then, right? Where also we are not sure where those apps were gonna be, but all of the CEOs and PMs and whatnot were super psyched and causing a lot of ruckus back then, right? So, and now we have something that is directly affecting us. And we know it's a wave that we need to serve but there's a lot of uncertainty and you're completely right So you need to talk about this uncertainty a lot Otherwise, you are just as you said you're like you're just losing people over that. Yeah, makes total sense André Neubauer: And you need to accept or you need to keep in mind there is this nice visualization where you have the entire population of the world and then people are using AI and then people are really using AI or working on AI. I think we are, like even our industry, we are at the very much forefront. If you look at large organizations who are still wondering, well, do we want to introduce co-pilot or not? or discussing that, we are far ahead of that. so I think it's just natural. the, first one, the first, the first people they, they experience and they are actually building up the patterns, the, the, the next companies then pick up on. so for me, actually totally natural process. yeah, I think we covered that now. And by the way, there's, yeah. Move on. Stefan: Yeah. No, sorry. I just wanted to, I just wanted to agree. Next turn of sense. Yeah. André Neubauer: Yeah. And then because I mentioned, I think there's also another thing to help the organization overcome that change. And I think we covered that in the talk with Bastian Boeck, so CTPO at the Getaway Group. Yeah. So I think they... Stefan: Get away. André Neubauer: reinvented or they redefined the roads. And I think that is also totally makes sense. So not just saying like you're a software engineer, but like come up with something very new, right? Where people can also start building up this identity. So they lost the identity. Now you need to give them the identity or you need to provide a new place, so to say. I like the idea, probably also a field which we will look into. So far, we haven't done it yet. Not sure if you have looked into that. replacing, not replacing, but let's say replacing the role description to be very precise, not people, replacing role description of a software engineer or combining that of a software engineer and the product manager towards a product builder. product engineer, think totally makes sense. this is, it maybe sounds simple, but I would not underestimate the possible impact it can have in organizations because people are just looking for that, right? They're saying, well, like, what is my responsibility? And you can make it implicit or explicit. I would always go for explicit. Probably you can change that on a weekly base. But once you're confident enough, I think that is one thing I would definitely look into. And we will do probably as well. Stefan: No, can't. Bastian is amazing. He always has a big theoretical framework at hand, so where he embeds his thinking, was observing him as a manager back then. There was the same kind of vibe and I agree that some of the rules need to be redefined. Right now I'm more at Especially the analyst roles are changing a lot, moving more to data products that are typically AI driven. So it's really... extreme how much we went away from just simple analytics platform reporting to having dynamic dashboards that are like where the technology is AI generated and then you have like it's really easy to switch between the data way easier than you can do in your typical I don't know Metabase or whatever you have right as a startup to observe it. So this day it's really clear. On the product side, I won't agree, so I'm not ready to redefine it yet, because what I'm observing a lot is that you actually need a lot more clarity right now on what you are building, because the engineers are becoming faster. And if you dilute the roles, then it's kind of the responsibility is becoming difficult. And oftentimes engineers, like, like not, every engineer is able to also cover the business side and engage with the stakeholders, do the user research or uncover it. You know what I mean? Like that's still, I think a rare, a rare trait. so this is where I'm not at. but I see the roles already changing. people taking lots of more responsibility, but oftentimes it's actually more responsibility within the same field. So more languages, more microservices at the same time, reviewing more PRs, having a better quality review. it's like right now I see more of the same actually, instead of the shift of roles, but I think both is happening and all of the companies are experimenting. like hell right now, right? So try those things. André Neubauer: Maybe also good reminder to invite Markus, Markus Andritschak again to maybe share, since he's working with a lot of companies, maybe he has already made up his mind and can share stuff. But I would also agree, for me, it's the compaction you see in the work, I think you also see in the roles. I think you will have less roles, especially when it comes to software engineering. Not sure. Stefan: Yeah. André Neubauer: about other disciplines, not an expert there, but for... Stefan: I think QA has a separate role like manual QA or somebody focusing only on... automatic this is like really difficult like I can see people building QA infrastructure still but because you're using AI tooling you just have way less people that need to do it right that need to set it up and then you have really good high quality testing in theory right because the technology isn't there yet in all of the cases but that's definitely something where I think that's really changing fast and there's a lot of competition also on the on the tooling side for that. It's just one example where I also directly see it. And I see designers pushing pull requests. It's the same kind of thing. Like if you have pure front end changes, I'm now seeing that as well. there's definitely stepping over the boundaries is already happening a lot. How the roll stand will change. Let's see. André Neubauer: maybe a good talk or good topic for another talk, the AI organization. So which roads are left next to the agent. So, but for today, I want to thank you for sharing your story. It was really enlightening. I think the separation of adoption and identification, so to say, is I think it's a very good one. Hopelessness could take some of that for their work. see. So thank you very much, Stefan, for having you. Stefan: Thank you, Andre. It was a pleasure.