#202 - The Rise of AI in Shaping the Future of Work - André Neubauer

[00:01:35] Introduction

Henry Suryawirawan: Hello, everyone. Welcome back to another new episode of the Tech Lead Journal podcast. It’s very exciting day today. I have André Neubauer today with me in the show. So today we are going to talk about something that is probably gonna be challenging a lot of perspectives about how we can use AI, especially in software engineering. So André recently gave a talk about this topic, right, in one of the conference, and I found that it’s very, very exciting, the way he actually came up with analysis and sharing his insights to the audience. So André, thank you so much for your time today. Looking forward for this conversation.

André Neubauer: Hi, Henry. Thanks for having me.

[00:02:11] Career Turning Points

Henry Suryawirawan: Right, André, I always love to start my conversation by inviting my guests to actually share a little bit from your career, any turning points that you think we all can learn from you.

André Neubauer: Yeah, well, like I can try to make it short. Um, so like working backwards, like I’m now 43. I’m in the tech industry for close to 20 years and started my career as an individual contributor, as most of us, and then made a career. And like over time, like I was always like super proud that it was really straight, right? Like so I started as an engineer, like I became a senior engineer, like became a team lead and a director. And I think, well, like if I could do things differently again, like I would make a change one or the other step to not make it so straight because I realized, and that happened to me, like let’s say during the last 10 years, like sometimes it’s also good to like go not the straight way, but like try out things left and right and learn a bit more.

May I share a bit, like I work for different companies, mainly in the startup scene. So as I mentioned, like started as a software engineer for ImmobilienScout, like most of your listeners probably don’t know, right? It’s a German startup. If you look for a flat, you probably know this. And then worked for DHL, like I think DHL people know, um, so a large logistic company, but like worked for the digital sector. And also during that time, like changed from one side to the other, actually late, during my late time at ImmobilienScout where I also met Stefan, right, who was in your show some episodes ago. Yeah, I think like one thing maybe to share there as a turning point, right? Like, so people are not aware, or we do not talk that much in our industry, like coming from, um, like individual contributor and becoming now like a leader, a manager, right? And what does this require, especially what you should focus on and what you should not focus on and like how you actually can grow. I think that’s the first larger step.

And then the second step is like, and that happened to me during my time at DHL, become a manager of manager. This is also again, very, very different. The focus, what is actually important, I think from my point of view now looking also back, is very, very different. Well, I made up my way like left DHL after some years, learned a lot. I think maybe also an advice here, like working for larger companies can be also like of interest, like, especially if you’re used to like the startup scene as I was back in the days. During my time at DHL, I really learned how to work in regard to like how to run like projects, how to structure problem statements, how to present things. There are also some good sites or some good reasons why to work for larger companies.

Well, anyway, left a large company and joined a startup again. And like work for a e-commerce company called Mister Spex, also mainly like operating in Europe. So selling eyewear. Worked there and grew the company from, well, like a startup more to a grownup. Was very, very interesting. Was a great experience also from a culture point of view. And then I decided, and that is, well, what I mentioned before, right? Like so left a bit like the straight line and decided like, let’s go for a different industry. And well decided for the FinTech environment, very different from e-commerce, was interesting. And maybe also to share that you can change things, maybe, and you may scared of that, right? Because you don’t know what you will expect. Maybe for that reason, it’s good to have some stability on the other side. What I mean by that, I joined smava. Um, I think it’s the oldest fintech we have in Germany. Like new industry, as I mentioned, but I connected with a friend, a very good friend of mine. He was the CPO. I joined as a CTO, which gave some stability. So it was like a good mixture of new things and old things. Well, did some time, spent some time there, merged with another company, also very interesting project.

And then, uh, left and joined Trusted Shops where I am now. Trusted Shops is again, I would call that a hidden champion. A B2B SaaS company now transitioning more to a platform company. And like, yeah, like again, mainly operating in the Europe and trying to bring trust into the digital space, mainly e-commerce.

Henry Suryawirawan: Thank you for sharing your story. So I think it’s really interesting that you brought up about this topic, right? So sometimes we should not just aim for a linear progress in our career, right? Sometimes, you can go left, right. Or sometimes even some people, you know, downgrade their level simply because maybe from being manager, they want to become an IC and just learn hands on coding or something like that.

So I think the key learning, at least for me, is that sometimes the nonlinear career can actually bring you much more growth and progress. And I think that looking back in your profile, right, actually you become like serial CTO for quite some time. So I think, uh, it’s really interesting in terms of perspectives.

André Neubauer: I think what is important is that it needs to be a conscious decision, right? Like so it needs to be explicit. So it’s actually like you, you are responsible for that. Yeah, and like back in the days I thought like, it’s like straight is the way to go. But I can tell you, doing some left and right stuff is like also a lot of fun. I’m like you mentioned, like I think I’m a CTO now for more than 10 years. And, uh, still, it’s not just that. Like I also consult companies or coach certain people, like I do some investments. Like it’s actually also that I think it’s good, right? Like if you’re curious, keen to learn new stuff, right? You also don’t need to leave like your, the stable part, right? You can also have fun left and right. You just need to still have some focus, maybe dedicate some more time. But this, at the end, it’s my approach. Doesn’t need to be the, what does need to be the approach for everyone, right?

[00:07:56] Giving a Talk on “The Role of AI in Future Workplaces”

Henry Suryawirawan: So let’s go now to the main topic of our discussion, right? So recently, you gave this talk titled, you know, “The Role of AI in Shaping Future Workplaces”. So I think we probably all the listeners here, we all heard about AI, we know about the impact AI has given us, especially in the last one year or so, right? So maybe tell us the background, how did you come up with this topic, right? And what made you wanting to present something from this?

André Neubauer: Well, like at the end, the idea of like generating code is like not new, right? So it exists for decades. Like even in my time at ImmobilienScout, like we thought about like code generation, right? Like so just modeling it and then like generating code. And I think that is, like I think it’s fair to think in that direction. And generative AI, or like at the end, like what is behind it? Like it’s the large language models. I think they bring in some new capabilities, right? Like capabilities with which haven’t been before there. And if you look at the development, at the pace, in which this area develops, it’s, I think from my point of view, just a question. Like when we are talking about like generative AI taking over larger parts or, at the beginning will not be large, but like certain parts of software engineering.

And it, as always, right, it starts small. It’s, um, less… As with every disruptive technology, it starts with some huge downsides, right? Also LLMs, they are not right so it really like at the end depends on the data set, which they have been trained on. But like this is just stuff which you can solve, right? And this is also the stuff which is getting solved. And so from my point of view, like I always get the concerns, right? But from a technology point of view, I think there is like it’s clear that will have an impact. It’s just like the question, like when and how much.

And for that reason, and I also shared that in the talk, and we probably will talk a bit more in depth in a second, but for that reason, like, if you say like it’s just a question of time and then the question of how much impact. What it lefts, it’s it will happen, right? And I think it’s a leadership responsibility to ensure your organization is ready for that and you taking the path. Because like still sure you could say like it’s not happening. It will never happen. But like honestly, maybe your competitors believe in that and maybe they create an advantage. So you should better be prepared, right? Instead of like having just gut feeling or whatever, and then disagreeing to that.

[00:10:30] What Drives the AI Advancements

Henry Suryawirawan: Yeah, you mentioned in the past, right? Maybe back then, 10, 20 years ago, we do have this kind of AI, you know, we do have some code generation, although kinda like, probably more primitive compared to now, right? But the recent advancement in multiple areas, things like, you know, LLM, the Generative AI part, right? And also the availability of GPUs as well and probably also a few other technologies, I think, make this kind of advancement really rapid, right? So I think maybe if you can summarize a little bit, what made the AI advancement so much different now?

André Neubauer: Yeah. Like you summarize it actually pretty well. And it’s also in the slides. You, maybe you, we, maybe you can link it then. But at the end, it’s like the three things, right? Like so, on the one side, you have like the massive improvements in GPUs because like they are used to do the, the matrix calculation and so on. So they more or less are built for that, right? So you need that in games, and like, you can also use that for model generation or like calculation.

And then like the second thing is more or less on the conceptual side. So these are the transformer models. I think it’s 2016, 2017, if I’m not mistaken, like where like some Google researchers came up with the transformer model. And like, based on that, like already back in the days, there have been LLMs, just the quality of LLMs have been very, very bad because like the capabilities on the compute side was not there, and that changed dramatically and it’s still changing dramatically. And this is also one of the reasons why I’m saying like, yes, like it’s maybe not perfect, but if you look at the degree, at the pace of GPUs are improving, like compute power being available or getting cheaper, it’s just a question of time until LLMs reach a state where you say it’s good enough for certain things, right? So I think this is more or less like really the foundation why we are talking about generative AI taking over certain parts at work.

Henry Suryawirawan: Yeah. And I’m not really into AI, but what I know, recent capabilities as well, things like RAG, right, so it makes, like, the LLM maybe less hallucinating. And also, you know, like, there are many assistants these days, agentic AI, and maybe, uh, there are so many other advancements that I’m not aware of. Yeah, so I think recently also like the bigger context window that, you know, some model are capable of. So I think all this definitely is gonna be changing a lot in terms of how we can use AI.

André Neubauer: Yeah. If you look back, I think, there’s this, I’m, from my point of view, there’s like step by step improvements, right? Like it’s in, one was this breakthrough GPT, um, I think it was late 2022, right? So it’s like, it’s already two years, wow! So I think all that, right, this breakthrough and like then also the availability of like compute power which improved then LLMs led to, I call that the rise of chatbots and copilots, right? So everything you see, like, for example, GitHub Copilot, I think it’s the well, super well known, right? That was actually, from my point of view, just the first phase. I like to call that the zero shot approach, um, and it’s actually, actually not something I claimed. It’s like a lot of people are using that.

What I mean by zero shot approach is like, you just use that one tool for a single iteration, right? You just like query it and then you take the result. And like you continue your work. So we’re talking about little integration, like, yes, it’s different if we talk about copilots, they sometimes they’re integrated like in your IDE, for example. But like they’re not deeply integrated in your entire workflow, right? And so it’s just like in your individual workflow, so to say.

And I think that is going to change next, you just mentioned. So while on the one side, you could say, we will, like LLMs with better and like will improve and improve. And I also mentioned in the beginning, at the end, like it’s just important what is the data set LLMs are trained on? And if you take like a generic LLM, right, for sure like the results will not be as great as they could be if you would train them on very specific data. And like very specific data is there, right? Again, if you look at GitHub, right? Like so the majority of open source projects are available there. So that’s a huge data set, which you can use to train LLMs.

And like you see this in the statistics. And there are enough benchmarks out there. Like the LLMs, like who have been like very specifically trained for that use case, right? Like, so the quality or the results they’re producing. It’s just really like super fast developing. And there are benchmarks which have been like very, very important, a year or two years ago. And like back in the days, like they have been sufficient because like no LLM was capable of really solving it. Like not in one pass, but also not in 10 passes or in 100 passes.

And this is more or less like now the standard. Um, I think the most known one is human eval. A test, I don’t know, 160, like, I think Python based 160 cases, like these kinds of tests, they can be solved by more or less every standard LLM nowadays. So think of that and like now use very specific LLMs. So I think this is one direction. You could go like having very specific LLMs. And like, just training it on more data, I think still we see like an improvement in using more data and having larger models.

And then the second thing, and by the way, like you also mentioned like context windows, I think that is a side effect. More or less like how much context you can give like an LLM, it is also improving. So like back in the days, like they have been very small, like context windows, maybe a page or two pages, like it’s always based on tokens. And now, we’re talking about millions of tokens, right? I think there is a beta phase of Gemini Pro. And this is also one of the, things are changing so fast, right? Like so when I prepared for the talk, like it was, I think, Gemini Pro 1.5 beta, which had a token window of a million, and there is some research going on for 10 million tokens. And like 10 million tokens is a lot, right? You can actually put your entire software project into that window. And it’s like in context window, and say like, okay, this is my application, help me building that. And I think all these kinds of developments that help like to produce better results.

And I think you also mentioned that, and I think that is the second large development is like not being like depending on this zero shot approach anymore, but like making use of generic AI in a different way. And what I mean by that is splitting work, so to say. Um you mentioned agentic AI. What it is about is like really thinking of what kind of steps do I need to take to solve a certain problem? And then you translate the problems into agents. So an agent is nothing else than, like an AI, right, who is solving that specific problem? And then it’s like, it’s handing over like the results to another agent and then to another agent. And what does this mean? You can actually break down the problem into very specific tasks.

And if you do this, complexity is decreasing. And with less complexity, you end up with better results. So that is like in a nutshell, right? Like so in practice, it’s more complicated. But if you see this and you also see like how companies or like open source projects, for example, apply this, you see like actually like small software teams, like where you have agents for research, for requirements research, where you see agents for software implementation, for testing, for security testing. And they all work together and like produce better results than this zero shot approach. And I think it’s very important to distinguish between the different streams, how gen AI is improving because it’s not just about the tools. It’s also how you use the tools.

Sorry for a long speech, but like I’m getting emotional, right? Um, like I’m, I’m very in, I’m, I’m more or less sold on the thing. And for sure, it also has some downsides, right? Like ethics and all this stuff, this needs to be solved. No question about it. But we should not talk down the possibilities this technology has.

Henry Suryawirawan: Yeah, so I personally haven’t really followed up on the agentic AI, you know, advancement. So I’ve heard here and there people talking about that, you know, using AI, interacting with each other. For example, also in security, you know, there is this so called the hacker AI versus one who actually creating the defense, right? So these kind of things are like…

André Neubauer: Black hat, white hat.

[00:18:54] The Levels of AI Advancement

Henry Suryawirawan: Yeah, which is getting more and more advanced. And I know in your talk, you also gave this kind of like overview, right? Maybe it’s from Sam Altman, right? Where you say that there are different levels of AI capability that maybe in the future we will reach there one day. So we have seen the first two like chatbots, the reasoning, right? So try to analyze certain documents and things like that. And now we are into this era of agents, right? Where maybe multiple AIs collaborating with each other. What are some of the other possibilities that maybe for futuristic people, they would like to understand where this is going?

André Neubauer: Yeah, happy to like… Think about it, right? Because like no one knows, right? You can’t connect the dots going forward, just looking backward. And maybe also to, to share that because like I also talk to people who have a different view in it. I really like to do this by the way. Because like I want to understand their concerns. If you’re like, I spend a lot of time, also a lot of my spare time to really go deep on these technologies to really understand like how can they impact my work, but also like the work in the company I work for. And for sure, right, if you just said, for example, like the second stage Sam Altman proposed is reasoning. Like for sure, we are not there yet, right? Like so we have reasoning or planning to a certain degree. But like, again, it’s not zero one. It comes in small portions and it’s also improving. So you could say for some of the problems we want to solve, like the reasoning we already have in place, it’s good enough. Is it like human quality? Is it with emotions? That, for sure not, right? But at the end, like you need to think it like you need to think backwards, right? Like so what do I want to achieve? And can I use gen AI for that purpose?

So just this, as a statement, because I get a lot of concerns in that direction. So you mentioned agentic AI, what can you do by that, right? Well, like we all, so far, we have only talked about software engineering, right? But like breaking down things into tasks, well, like you can do this everywhere, right? It’s not limited to software engineering. It’s, you could also apply this to like, I don’t know, like your operations department, right? And I think you also, like, this is also not new, having these kinds of chatbots. But like extending the capabilities of a chatbot and having some agents in place to take over the request from a customer talking to a chatbot, and then like changing things based on that request, I think that is a no brainer.

And if you like continue thinking about that there are probably a lot of things, you can automate in that way. And the way I look at it is it’s for me, it’s something like UiPath, if you, if you know that technology or that software. Like it was a hype technology. It’s called like the field was called RPA. robotic process automation. Absolutely. So like, I think it was very famous UiPath, I don’t know, five, six, seven years ago. And for me, it’s like UiPath on steroids, right? Because back in the days, like how you automate it. Like it’s really like click there, click there. Now it’s more or less you can give it a context, right? What do I want to achieve, like solve that problem. So it’s way more powerful.

But I think I, well, I’m also, again, like, I’m not talking about like generative AI or like agentic AI replacing all our work. Maybe, in the future, no one knows like how long it will take. And I think also Stefan like has a very strong opinion, like that coding will go away. Not sure whether, maybe at a certain point. But it is very blurry when it could happen. But I think what is very, very achievable is like improving certain, or like, yeah, replacing or automating certain tasks in your daily work, right? Like, so I think if we can make it simpler, why not, right? Especially if it’s about the boring work. And one of the things, well, and honestly, let’s talk about that as well, like I think there are two challenges there. There will be also one large challenge for existing companies. I mentioned before, right? You may not agree to the capabilities of Gen AI, but your competitors will. So they may end up with a like advancement or advantages, but also like new companies, they do not have the legacy, right? So if, and this, I find very, very interesting.

I think there’s also a term for that called solopreneurship, where a single person or a very, very small team is capable now running large organizations. And if you think about that, like you can achieve that productivity gain, like from two directions. Like one, you have a large company and then you need to like automate certain functions. And like, this can be painful because you need to deal with a lot of concerns. But on the other way, right, as a founder, right, you may not require so many people because like from the very beginning on, you say like, let’s make it a habit, and use technology to handle certain cases.

And if you want to understand where this could lead to, I, I mentioned also on the slides is like, just search for the Dreamforce 24 keynote. So it’s like the largest event by Salesforce where they show their product improvements. And this year it was all around agentic AI and how like you can use Salesforce to automate your work. And I’m not saying like you replace 100%, but again, right? Like certain steps in the chain, in the operational work can be automized, right? And this can become a huge advantage, especially for startups. And that could also lead to some interesting situations where like very, very small companies challenge like the dinosaurs, so to say.

Henry Suryawirawan: Yeah, so I think one key takeaway for me when I listened to you, right? We should not be attached to our own habits, right? So what we are used to in the past, it may work, it may still work now, but the-advancement I think we should keep an eye on, and maybe try a few tools here and there, try to automate some of your tasks, try to learn what could be the things that can gain your productivity. And like you said, right, even if you don’t adopt it, other people may adopt it, right, and your competitors or maybe new disruptors. So one day, you’ll get affected anyway. So I think thanks for reminding that.

[00:25:01] AI in Software Engineering

Henry Suryawirawan: But speaking about that, right, let’s maybe focus a little bit now on the software engineering, development world, right? I know in the news there are so many news happening saying that, okay, developers job may go away or may get reduced a lot. Junior developer may not be required anymore. There are a lot of layoffs as well. And there are tasks that can be solved by AI independently without human interaction. Maybe from your point of view, what are some of the most interesting news that you think could be a breakthrough for us software engineers to actually learn about AI?

André Neubauer: Yeah. Well, like, I think the most boring use cases really generating code. And at the end, it’s also very hard to say like this is solved or not, because it’s at the end really depends on the domain you’re working at, right? But what I find interesting is like by getting away with some boring stuff. I think documentation is one of the things which you can, where you can use generative AI, to a very large degree. Um, maybe also analyzing stuff. I told you context windows are so large nowadays. You could basically put as long as like you do not, we don’t talk about the monolithic application. You could throw in the application into like the LLM and then just query the stuff. I found that very interesting to also get a better understanding how software is functioning.

Like if you think of like the ratio between us or software engineers writing and reading code, right? It’s like still, I don’t know, 10, like one to 10 or one to eight, doesn’t matter. But so like the majority of our time, we are reading code. So if we can improve here by having someone like who’s explaining, like actually you have some someone like something like an expert, right? Like so the person who wrote that stuff and can help you understanding it. So I think that is from my point of view, more or less like the hidden gems, so to say.

[00:26:53] Concerns on Tech Debt and Issues

Henry Suryawirawan: Yeah. One thing that in my experience when using AI as well, I mean, it’s not 100 percent perfect, right? Sometimes it could throw a suggestion which are totally wrong. Sometimes it could be partially wrong and things like that. Obviously, one of the concern here is about, you know, the amount of quality that these kind of code generation produce, right? So sometimes it could be buggy, sometimes it could be security issues, right? Sometimes it could even lead you to making a big mistake if nobody’s reviewing and you just push it into production straight away. What are your thoughts as a CTO, probably, right? Looking at your team, for example, they are adopting AI. Would you have concern about tech debt and security issues?

André Neubauer: Yeah. Well, I think on the first point, what you mentioned at the end, it’s still called co-pilot, right? So there’s a reason why it’s called co-pilot and not pilot. I think also the software vendors are very much aware, like that it’s not at a state where you can say like, just fire and forget, right? Like so generate it, put it in production, it will work. So you need to be careful. Like there’s no question about it.

What I like a lot, and I would, it would be also an advice. Like you mentioned it as well some minutes ago. Try, try, try. You need to try it, right? Like, so I can tell like for my environment, the company I work for, like it works, but it does not mean it works for every company because every domain has its also non functional requirements, right?

So you may operate on a like very special programming languages. You may have some like real time requirements. All these stuff, right? Like so they have an impact on the decision, how much you can use. But what I would say is like give it a try and maybe also to share that. Like once a month, we have an entire day at Trusted Shops, where people, we call this self education Friday. So it’s always the last Friday in the month where people can spend an entire day and like to learn new stuff. So we encourage them, for example, to give it a try. Because at the end, like this is telling you whether it works for your use case or not.

And then also not just rely on one technology, but also other technologies. So not just like saying, okay, I use GitHub Copilot, like Gen AI is solved for me. It’s not right. So you also should think is like, how you can make use of it across or beyond that. For example, think of like also how you can solve certain problems like you mentioned, hallucinations. or like certain other aspects. So just thinking about it, you could also say like, you do code review by an agent, or you have like five code agents, five code review agents. And like everyone has a very special role and should focus on certain aspects.

I think it’s, at the end, it boils down to like you need to put this into your context. And then it’s like dedicating time is what I would do as a first step. And then as a second step, you need to create, I mentioned it before, from my point of view, it’s more a mindset shift than a technology shift. You need to create this kind of success stories in an organization, right? So I would not apply it to a large degree in your most critical project, but maybe there’s a small project where you can sacrifice whether it’s when it’s a week late or like you face some, some issues in engineering. Well, like, let’s give it a try, like give the team the freedom to try it out.

And like, again, not just learn like outside software engineering or like outside like production software engineering, how you can make use of it, but also like in your day to day work. And then the last step is like applying it everywhere, right? And again, you need to think of it like short term, mid term, long term. So I would not advise any company to say like, let’s apply it everywhere. This will end in chaos.

Henry Suryawirawan: Yeah, so thanks for reminding. I think the technology has not reached to become like the pilots, right? Even worse, the agentic pilots where they just automate everything and do the task. So I think it’s still up to us, the human, to actually leverage on those technology as the assistants, as the co-pilots, right? And maybe even like to give us some new skills that we didn’t have before. Like, for example, you mentioned about reviews, right? So sometimes code reviews could be a security aspect or some new language that we are not familiar with. So I think all these, you can actually benefit a lot as well, not just to automate the task away. But you can use that to improve yourselves.

[00:31:11] Impact of AI to Organizations

Henry Suryawirawan: So what I like about the topics that you presented back then, right, is actually the impact to organizations. So maybe let’s just go there, because I think in many organizations, people are scrambling, like how can they apply this technology? How can they assure people can still have a kind of like comfort with their job? And how can they become more productive as well? So maybe let’s start with the first one, which I find really, really interesting for organization to think about, right? You mentioned that now with the availability of AI, the basic understanding is becoming a basic skill. Kind of like you don’t need someone maybe who has, I don’t know, like a more boring menial tasks that can also be automated by AI straight away. So tell us what’s the impact for organization for this kind of thing? Do you think that all those, I don’t know, like lower level jobs will go away?

André Neubauer: Yeah, this is a good question. Like, I don’t wanna scare people, but I think you, again, I think it boils down to this very specific use case. And I think you just need to give it a try. But most likely, you will be surprised how much you can automate. What does it mean for me? Or like what does it mean for software engineering? I think software engineering will be required, like actually to a larger degree going forward. Because someone needs to write all the agents, needs to do all the automation. And honestly, like also someone, even though, right, like so at a certain point, like co-pilots become pilots. Like someone needs to oversee this, right? Like, so it’s like, still there is work to be done. It’s just like a different kind of role.

And if you think about that, like also in our industry and like software engineering, like when I started software engineering, we still deployed on hardware, right? Like so not containerization, no cloud. And like that, even back in the days, right? Like it felt natural. And then all these new, well, let’s say new capabilities, right, improvements. They came up and people have been also concerned. I very much clearly remember when someone told me, right, there’s someone like who is deploying the software, like several times a day, it was like, huh? How? Why? Right?

And so I think like the fact that our roles are changing is like natural, because our industry, our craft is like still very young. And it’s, I think, you need to accept that. The fact that you don’t like maybe code that much, like produce or implement that much does not mean that you are not needed anymore. It’s, I think, it’s just like a bit differently. And in regard to what does this mean for organization? I think, and I’m really happy that it happens, we will most likely see software engineering teams everywhere in the organization. Because like, if you start having this mindset, well, like you want to make use of that productivity gain, right? Like so someone needs to do this.

And then back in the days, also like companies I worked for, like software engineering was mainly there when it comes to really product implementation, right? So the product, the company is selling, but it’s like, was really there where like the operation work happens, right, there, you have more or less like the enterprise systems. And they will also stay, right? So Salesforce and all this stuff, they will stay. But like there is so much stuff which needs to be automated. And I really think like we will see software teams there to like, help automating it.

[00:34:34] Smaller and Leaner Teams

Henry Suryawirawan: Yeah. So I think that’s also another unique insight. You mentioned that software teams will be a lot more, right? So maybe in almost all the teams, you will have some capabilities for software engineers in some form or shape, right? It could be, you know, like a more integrating stuff, right? And this leads to, you know, what you mentioned as well, that there will be a lot more high performance teams which are leaner, right, and smaller in size. So tell us what’s the impact of these to organizations.

André Neubauer: Yeah. Like there is an interesting blog post, um, like what AI or Generative AI could mean to agile software development. And at the moment, we like, what is an ideal team, right? You want to have a cross functional team. You want to ensure like it’s not too large, but it’s also should not be too small. So if someone is sick, someone’s on vacation, still someone should be there to maintain the stuff. And Gen AI is changing that thing, right? Because like you could have like all skills in the team, right? Like so the skills you just acquire because like the next agent is just a fingertip away. Like it’s just like a question about compute power. So therefore I would expect, and this is also what the article is saying. Like teams will be smaller, right? And therefore also work will be different. And teams being smaller does not mean like the company is now letting go people. But you could use that, you could use the people to form new teams, right? So you end up with more capacity, and this you can use to like slowly, but steady, right, rework your processes in the organization.

Henry Suryawirawan: Yeah, so I think that’s a very interesting thing, right? More teams will be formed and you, by the result of that, right, you will be able to do a lot more things, right? So maybe not just productivity, but you may expand your area as well. Not just doing certain capabilities that you have, you could expand more. And plus, I think you mentioned also like with this capability of AI, right, you could do a lot more innovation, research, kind of like analyzing new things, which you didn’t know before, right? Much, much faster, because in the past, I still remember we all do a Google search, right? But sometimes it’s like very difficult to kind of like summarize, because you have to navigate one link over to the other and over to the other, but now it’s like, you know, you could just ask questions, and they will analyze, summarize, give you citations as well. Although sometimes it’s wrong, but I think it’s better than, you know, like doing it yourself, right?

André Neubauer: Absolutely. And then thanks also for reminding me, like I talked about a lot about like productivity. But like, you’re absolutely right. It’s like, it’s also about like research and innovation. Like you literally have time to care about the most important things. I think you can boil it down to that.

[00:37:15] The Rise of Solopreneurship

Henry Suryawirawan: The other thing that, uh, you mentioned in the beginning as well is the rise of solopreneurship. I think for some people, this is also something that they aspire to do. You know, creating small business where they can create, I don’t know, like profit and income for themselves, small family. So maybe in your point of view, how can we leverage this to become more solopreneurs?

André Neubauer: Well, like at the end, I think solopreneurs will most likely be tech people, because you need to have like this awareness, this, like this kind of thing exists, right? Like, so solopreneurship is, I think like there are a lot of solopreneurs out there, but like there is a certain limitation, right? Like so this is a limitation when your time available is coming to an end. Given that there is some, also some natural limitation about like the size and, well, like the revenue, for example. The revenue capability of a company, because like there’s some kind of correlation, right? Like so you need to invest time to reach to more revenue.

But if you apply now, like Gen AI or agentic AI to that situation, well, like you could make it a habit, as I mentioned in the beginning, right? You could say, like, from the very beginning on. Well, let’s clearly think about that. And like whenever there is something which is coming, which we continuously need to do, right? Or like which we want to discover, like always have a mind, like how can I make use of Gen AI to like improve that? Or maybe also to take over.

And, I think like the way also software gets built, like also internal software, it will be very, very, very different. So like the question, like one question, like for example, why do we need Salesforce, right? Like so at the end, we need a database and maybe in like an application layer for some logic, but like maybe also front ends will disappear. And I think that if you think that through, it can be really wired, what does this mean, right? I mentioned it before, like companies, like large companies, maybe like not becoming solopreneurs anymore, but like becoming more automated. So they will stay with a certain complexity.

But on the other side, like, so I think you will see companies just making use of that concept who will look very, very different internally, right? And then like hard for me to like think that through, because like, it’s really like creating some conflicts, so to say in my head. But, um, like, I think it’s hard to foresee how like these AI or agentic AI driven solopreneurships will look like. But I could imagine like it’s very, very different from what we see nowadays with some applications here and there and some automation in between. I think if you started from, if you applied it from scratch, it will look very, very different.

And maybe also to share that thought, also I did this longer discussion with someone who like is kicked off some startups, you should not mix that up. Like was really being like innovative and dynamic in the beginning, right? I’m not saying like, it needs to be excellent from the very beginning. It’s more or less like being aware that you should not think of like the next or the first employee, but you should think of like, what kind of like, for example, agents you require to get some stuff done. So like the mindset should not change in regard to disruptive, innovative, and like, also like maybe sacrifice some quality in the beginning. But more or less like how you apply that or how you achieved it.

Henry Suryawirawan: Yeah, personally, while I’m doing this podcast, I’ve been leveraging a lot more AI tools as well. Yeah, sometimes, it’s really amazing, like, what are the capabilities available out there for cheap as well. Like, you just subscribe monthly, give it a try. If you don’t like it, you can stop, right? I think this opens up a lot of doors, opportunities for, you know, a single person or small team to actually achieve a lot more, right? In the past, probably you need to leverage on, I don’t know, like the freelancers. You know, find it maybe sometimes the person is good, sometimes is not. But now you can actually do many of those things by yourself, right? I think definitely is a door openers for some people who are, you know, into creating side business or you want to do content generation or whatever that is, right?

[00:41:32] Getting People Onboard to AI

Henry Suryawirawan: So I think as a leader, as a CTO, definitely, you know, one aspect is to figure out how to use AI within your company, to make more productivity and, you know, gain more things, right? But the other aspect is you have a lot of people who are concerned as well. Maybe they also lack the capability to use the AI. And I think in your talk you remind people that actually in organization, it’s still pretty much a people business, right, rather than just focusing on technology. So tell us your view about this, actually.

André Neubauer: Yeah. I mentioned it partly at the end. I like to say like tech is a people business. So I think you can disagree like, on the degree how much generative AI will help you getting some stuff done. But if you look below the surface, why did like people have different views I found, I would say it’s basically not based on like facts, but more on emotions. So more on the soft side. So you need to, like, first of all, you need to understand that and then you need to address it. So there’s no way to say like, well like I disagree or don’t believe or like shut up or whatever, right? Like, so it’s, from my point of view, it’s like change management as you would apply it, like if you change your organization, right?

Like so you need to understand the people, you should listen to them. You should also like openly talk about their concerns, right? You may should also address it. I’m like, as I mentioned in the beginning, so I’m like, I’m sold on that thing. So I’m also partly get blind on certain areas. So it’s a team effort to also make that change. Um, I just would say like, just based on some like concerns not doing it, I think it’s the wrong way. Again, here, like you also win as a team, right? It’s not about just someone being right. It’s more about like addressing this change. Like being open, being honest, listen, maybe also adjust stuff. But then at the end, like I mentioned it, also like you mentioned as well, like try, try, try. Turn a theory into practice and like in small steps, implement it in the organization.

Henry Suryawirawan: Yeah, so I think definitely the people concerned, right, you need to hear about it, not just focusing on the, you know…

André Neubauer: Yeah, absolutely.

Henry Suryawirawan: … so-called the bottom line, the profit, the numbers that you need to save, right, the number of people you need to reduce. I think many of leaders these days are thinking more on that front, right, but actually the people aspect should never be neglected.

André Neubauer: Absolutely, absolutely. And that you mentioned like people letting go, like I would be definitely on the other side. Like, I think, I like to say in Germany, we say like IT is Mängelverwaltung. So there’s always more to do than you have capacity in your organization. Actually like it’s an interesting development because you can get more stuff done with the same amount of people. And it, isn’t it cool, like if people can like also discover new fields, right? And we can have an impact on the organization, can deliver value like internally as externally. I found it fascinating. For me, it’s not a question on like how much people I can let go. Like for sure not.

[00:44:40] Leadership Measures for Adopting AI

Henry Suryawirawan: Right. So thanks for the reminder, right? So what are the things we could do more, right? I think I love that one. So I think maybe a little bit summary as well in your presentation you talk about there are short term things we can do, medium term we can do, long term that we could do something as well. Maybe one each like what can the leaders have in terms of point of view. What are the things that they could do in the, in this progression, right? Because obviously in the next few months things may change again. But something to think about.

André Neubauer: Yeah. Like first of all, well, like I don’t want to repeat stuff I already mentioned. So like this, like you need to have the safe space to give it a try, right? But like I mentioned it so much so often so I will skip that, but like don’t forget about that. I think like you should have a discussion on real world facts, not on theory.

I think what is cool is to have these kind of like small projects, which you can run in a company to give certain things a try, which have a low risk profile, and then like turning it into a success story. And like, also like really carefully invest in these kinds of things, right? Like so when we implement certain functionality, at the moment, like our, like, well, not our first AI or Gen AI features, but like still some teams do not have the experience yet, right? Like, so staff the right people. Also maybe get external people in the project to really ensure like this is a safe environment for the people, you have the right skills and it will be a success story, right? Like, so you really need to ensure it is a success.

Second is more for the midterm thing is getting the buy in from the organization. I don’t know, what’s your experience, but like I still find it fascinating if I see non technical people like handling technology. Like if you think about that, right, like, so I think they hardly can imagine what’s going on at the moment with Generative AI, right? It’s, for them, it’s probably pure magic. And like creating this, like, not that understanding what an LLM is, but like really like try to bridge, right? Like so the gap, I think that is important to also get a buy in and like maybe also partly develop an understanding where they can use that.

Well, like maybe then that brings me to a long term thing, or maybe a midterm to long term thing is bring in like, if in case you do not have that right, bring in that experience into these departments in the organization. So still a lot of departments, they are, like they’re non tech. It’s also like the case at Trusted Shops for some departments. They’re just like, they have their tools, like they have their solid processes. Like it’s working for a longer time. So that is like it’s just working without having technical people in the department. But I think if you want to change stuff, right, you need to slowly, but steady, like build the right foundation for it. And what I mean by that is like also making sure the organization is staffed right. Because like, as I mentioned also before, you don’t want to end up in a chaotic system where people just try out all this stuff. And like you need to maintain a zoo of different technologies.

So maybe there’s also something for short or midterm maybe last because it is maybe also like not just the enthusiastic part, but also you need to be, you need to handle it carefully. So having some kind of AI policy, if you want to call it like that. So something which is giving some boundaries, providing some guidelines to clearly say what’s in, what’s not.

Because at the end of the day, right, you could end with a lot of like processes or like projects where you want to make use of it, but do not have the skills. And then you end up in a lose-lose situation, right? Like so and then, it’s more or less you bury the topic, and like people are more concerned about it than before. So for that reason, maybe also having some guidance is probably helpful and also something I would advise to do in the beginning.

Henry Suryawirawan: Yeah, and obviously the never ending topics for the long longer term is about the ethics of AI, you know, the security, privacy of your data, right? And those kind of stuff definitely is always in the back of the mind of the leaders. So thank you so much for this opportunity. I feel kind of like, I don’t know, encouraged after this talk to actually explore even more about AI, right? And I think the listeners here, when you hear about this conversation, right, I think you should feel encouraged as well. Because the possibilities are definitely, I don’t know, it could be like indefinite, right? Maybe in the next one year or so we look back and it seems like we progress a lot rather than, you know, taking the stance that, I don’t want to change, I don’t want to give it a try. So definitely that might not be a good strategy. So thank you so much for opening up new perspectives for us. And I’m sure we got to leverage a lot of AI in the future.

[00:49:34] 3 Tech Lead Wisdom

Henry Suryawirawan: So André, as the custom in my show, I would like to ask you the last question, which I call the three technical leadership wisdom. So if you can think of it just like an advice that you want to give the listeners, maybe you can share your version as well.

André Neubauer: Yeah. Like three things, like hard. One thing I mentioned already, pretty much believe in that tech is a people business, even in, in an AI world, right? People are running the show, right? They may do not code that much in the future or like less. But like, this is still like tech is a creative business. Um, so you need to care about the people. So this is the first thing.

And second is, for me, the world out there is complicated, right? If, well like, complicated and actually also complex. Sometimes it’s easy to like, or if you want to understand situations better, I think having some kind of mental models in place, which help you to really like structure things, like getting away from the complexity, unnecessary complexity is helpful. So I spend a lot of time finding these meta models because it helps me to like do my job in a better way. Like to just mention one or two, like we talked about like short term, mid term, long term thinking. I think this is one thing, right? Like so not just thinking about like this one solution, but like what is also the path towards it? Another thing could be like how you like break down the problem. Like I like to use situation, complication, resolution framework, these kind of things. There will be my second.

And the third is like, this is a boring one, or like you, like people probably mentioned that, over and over again is never stop learning. Like our industry is so, so young. And this is also like maps to the AI topic, right? You should be open. You should be open to new stuff. You can disagree, but you should disagree only like if you have the real facts and made certain experience, not just like by gut feeling. So always learning, always be open. Um, I think that would be my third advice to survive in that game, to survive in that game.

Henry Suryawirawan: Yeah, not just the industry is young, but I think the capability are also kind of like a lot of things are new, right? Coming all the time, you know, every day. So I think that makes it even tougher, right? So I think thanks for reminding that.

So André, if people love this conversation, they want to find you to talk a lot more about AI or maybe other things as well, is there a place where they can find you online?

André Neubauer: Yeah, sure. Like since there is so much going on, I also decided to focus a bit more. So you mainly can find me at LinkedIn. And you also like, I also every now and then publish an article Substack. But you will find us all at LinkedIn. So LinkedIn is the right place to connect.

Henry Suryawirawan: Yeah, so André’s newsletter is Tech Advisor. So sometimes he wrote stuff about, you know, CTO stuff, you know, engineering leadership. So make sure to check that out as well. So thank you so much for your time, André. So I’m sure that we all can learn a thing or two about AI today. So thanks again for that.

André Neubauer: Thanks for having me and hope I could share a bit my enthusiasm.

– End –