Fehlt Deutschland der Mut für das KI-Zeitalter, Herr Dettmers?

Hello on Saturday, also this week, about our CEO edition of Table Today. It's nice that you're here. Today we're talking about an interesting company, more specifically with the CEO of a really interesting company, called Stepstone. The CEO is called Sebastian Detmers. The labor market has changed a lot in Germany and also the way we find our jobs. In the past, when the older people remember it, the weekly expenses of the newspaper were always so big that they hardly fit into the mailboxes. And the main reason was that the workshops had blown up the newspapers like this. Also a bit of the housing market, but above all the workshops were searched and offers were placed there. Meanwhile, this business is of course running digitally and Stepstone is something like the Amazon of Jobbörse. The labor market is of course a huge topic in Germany and the problem is no longer that we used to have too many unemployed, but the problem is much more that there are not enough qualified workers and there are no workers. And if there is a lack of manpower, then this is a main obstacle for the growth in this country, in which we all wait to finance this, where we talked about it yesterday in our podcast, we talked about the "Torre Gesundheitssystem". But there is hope and that is what Sebastian Detmers of Stepstone says and hope is based on artificial intelligence. Because the more processes are automated, the more productive a company can be without having to deal with more people. The labor market is changing and the reason for this is that Stepstone has placed the issue of AI at the very beginning in the priority scale. The editorial director of our CEO Tabels Alex Hoffmann talked to Sebastian Detmers about how AI will continue to change the labor market and what German companies have to do to be able to do so in the future. Hello Sebastian, welcome to the Tabel Podcast Studio. Thank you very much. Sebastian, in your book, the great worklessness, not unemployment, worklessness, you write from the lack of labor force in Germany that will lead to massive changes in the working world. The book is 2022. Since then, a lot has changed. AI has become a subject for all managers. The unemployment numbers have gone up again, from good over 2, now almost over 3 million people in Germany who don't have any work. The book was very well known back then, because you already made a few announcements. At least a loan of 20 euros, for example, comes in the book. Are the thesis still in your book today or are there even more sharp measures necessary? Yes, I also thought about it. I looked into the book again and asked myself what it actually was. And in many things, I don't think I just said well before what happened, but I was almost too conservative in the drastic development. Because the book really deals with two things. One is the demographic, i.e. the birth rate and the development of the population in Germany and all over the world. But also with the question of productivity. So what actually creates the people who then still work? Because there we have a phenomenon that productivity has gone back over the years or decades and we have a stagnating or even decreasing productivity per head. I wrote it like this. If a stagnating or decreasing population rate hits a stagnating productivity, we may experience decades of recession and I thought it was a bit controversial at the time. It could be a century of industrial revolution, a century of industrial step back. Of course, I didn't say that as a prognosis. I said that it was a danger if we don't control it. And we have to ask ourselves how do we function as a society with an now-alternate, soon-spirited population and how do we create productivity again? About three years later, we experienced a 22-year economic rise, we stick to this decline. The population continues to stagnate. We realize that we will have big problems in the future on the labor market when we look at what is happening there and what is happening at the top, at the pension age. And we simply don't get the issue of productivity in the attack. But we still have so many tools. Now there is still AI. What is it that makes us so unproductive? That is a question about which scientists break down. Interestingly, especially outside Germany, they call it productivity paradox or puzzle. There is not a single big explanation, but there are many small ones. First of all, the biggest part of the human history was actually no progress. Productivity increase, as we know it from the last decades, is a total exception. And what happens first? Is there still enough innovation? Can we talk about that? I think the second question is, is this innovation adapted quickly enough? We have started with all great innovations, starting with a steam engine, telephone, internet. Smartphones, as we have always seen, take about a generation, until great innovation actually arrives in the wide range. Alternative societies have longer adaptation cycles than new technologies. Is that a problem? Then the price is very low in Germany, almost at the lowest level. That is a disbalance that leads to less innovation from young companies and startups. I have mentioned a few reasons. There are other reasons. There is not a single reason for it. The question you have to ask as a country, as a society, is where will the next productivity turbo come from? Just like it was always the case in the past. With the automation of the industry, with the introduction of computers, how do we do it? For example, in such a dirty analogy, how do we manage the work for which we need eight hours today, tomorrow maybe four or two hours to do it? Those were the progressments that our ancestors experienced, and we have to reach that again. And is AI the answer to the question? AI is, I think, the most important answer that we have on this topic today. I hardly see any other answer, because the economy is part of the service sector. And the automation industry has long been hacked, just like in agriculture. There, machines are used everywhere. In a modern factory, we no longer see many people. But we see these people in the office. And there, they still work to a large extent, repetitively, they do the same every day. And of course, we don't need a robot or a machine, but we need algorithms, so artificial intelligence. And the core question will now be, do we really set artificial intelligence so that it automates work, so that it makes us more productive? That's true, and we rely on other countries and other companies to use the productivity gains that are certainly possible with artificial intelligence. There are several studies that find fewer and fewer employees and increase their jobs directly. You probably have the largest job data bank in Europe. Do you see that in your data too? Yes, we see that in our data. We see that the entry-level jobs are currently 45 percent, 45 percent under five years' average. Of course, we also have these super strong years in 2022, where we have experienced this enormous job boom. But we observe that. But you have to make a distinction. There are two trends that overlap there. One is the conjecture. We stick to the recession. Because of that, companies are generally less involved and are currently suffering from entry-level jobs. The second is actually the question, how do companies prepare themselves especially before the age of artificial intelligence? Because it's not just that companies take jobs or have a conjecture crisis, but we see differences there. Especially the jobs are affected where in the future I can fully automate work with artificial intelligence. So what is that, for example? That's bookkeeping, that's jobs in staff, jobs in marketing. Things that I can actually automate with artificial intelligence today, we don't observe where artificial intelligence is not automated. In the handwork, in the care. In this case, it's a differentiated picture and it fits well to our assessment of where artificial intelligence is usually changed. But I also heard that many companies don't prepare well enough for this situation. That can be read in simple numbers. Because in the end it's not about whether I use artificial intelligence or not, but it's about being productive as a company. Whether I use it more or more with the same number of employees or if I can't use it I can do that with fewer people. That's the reality. That was always a progress. And that has also contributed to the prosperity we experience in this country today. Let's take a look at it. Productivity stagnates. We have been working in the city since 2019, prosperity even since 2018. So yes, we see a lot of exercises, but in the broad mass it doesn't lead to the prosperity we would have to do. And I think there are many different roles. First of all, it just takes a while until I have implemented these algorithms and really made sure that they make me more productive. Second of all, we actually go with the will to, I say it in a more aggressive way to this point, we are ready to think in a completely different way in private economy but also in public administration. So I would like to say, is it about giving a employee a tool that makes my everyday life a little easier or is it really about automating the whole business process? And so, we still have a lot of work to do. Stepstone is a tech company. You should go as a good example. What are the big AI developments that you are currently working on that can make you more productive? Stepstone is a platform where we can form people with people who want to work together. And with this, we form a deep human decision on our platform. So if we would now decide to work with each other, then of course it's about qualifications. What did I do before? Which experiences do I bring with me? But we actually also fit together. We can work together. And that is a complex decision. We notice, even if we have to make this decision, how much we have to deal with it, it is not easy to form a data model. And here, artificial intelligence helps to understand people in all their versatility, to understand how different jobs are. Of course, it makes a big difference whether I am a doctor, a caretaker or if there is a job or an opportunity where it is only about that I have a driver's license and an employment permit. And in all this versatility, artificial intelligence simply manages to answer questions that we cannot answer humanly at all. So what does this mean in practice now? The artificial intelligence will be a big matchmaker. We will soon have a tool, a small agent in our pocket who actually keeps an eye on us permanently. There is a better job there. Which always means better for you. More money, less pendulum, maybe even shorter work, whatever it is, being more attractive to your employer and which also helps companies permanently to find the right employees. Because it is also an irreplaceable puzzle that we are constantly trying to solve there with tens of millions of employees and millions of jobs and hundreds of thousands of companies. And artificial intelligence will simply know this puzzle much, much better. And with that, it is also very likely to contribute that the economy works better and that people are happier in their jobs. What does "bald" mean? What you said yourself, it comes pretty quickly. If you say we have the "bald" in our pocket, can you give us a timeline? First of all, we already have the "bald" in our pocket. We have built long-term algorithms that actually train with tens of millions of jobs in historical and user behavior to prevent people from doing the right job. So today we send our users with a push notification a job on smartphone if we believe or if the algorithm thinks that this is the right job. We can tell our users how good you actually fit on this job and how high is the probability that you actually get this job. And of course it also works for the employer. In the past, I published a statement and waited for someone to comment. Maybe no one commented. Today, I publish a statement and I immediately get selected candidates with the help of artificial intelligence. They can invite them to the interview. They can invite them to apply. They can directly address them. Those are the prerequisites until the last step that the whole thing really runs autonomously. That I actually only press "yes" and the intelligence tells me that I found a suitable job and I should get in touch with the recruiter. Of course, this will not lead to people being automated. At the end of this process, at least in most cases, you should still get to know a personal conversation. But this is much, much faster and also much more suitable so that we don't have to waste time with the sent out applications on not suitable jobs, not at all suitable. The AI is becoming a big matchmaker and we have built the algorithms. Now we have to make sure that the companies adapt to their processes. Because of course we realize that a lot is still going on like in the last 30-40 years. But I think the profit of employees is probably the biggest technical revolution in their history. But we also know that artificial intelligence always brings a bias. How do you deal with it? I will briefly mention the EU AI Act, the regulation of artificial intelligence on the European level, which many of our applications described as a high risk system. This is the highest form of regulation. Why? Of course, A is about very important processes, i.e. the selection of employees. And there is this issue of the injustice of the bias. But it is not made by machines, it is made by people. When we talk about discrimination in the working world, it is not about discriminating machines or algorithms, it is about discriminating people. That means, when we assume that the machine advantages have biases, it is not the concern that they really have, but that these machines learn from our bad decision-making. And the task is not to regulate artificial intelligence per se, but to ensure that they do not look too much at the bad decision-making of people. Of course, we can say that this is a big risk. In reality, this is a huge chance. Artificial intelligence is, of course, an ability to prevent these biases and to make us aware of the biases that we have in our decision-making. If you go there and say, "I'm charging 100 potential applicants again and again the same according to certain demographic details, people to introduce a proposal or put them in, then artificial intelligence will be found at some point and say, "You don't know the best decisions for your company." As a company, and this is very important to understand in a market, we have no interest in letting biases in our algorithm. We earn money to help our users, help all users find a job and help companies to implement the best possible. That's why we invest a lot in it. We have ethical AI teams to find ways to avoid these biases and also to prevent the algorithm from finding a way around these biases, which is made by people. First of all, to somehow build into the algorithms. In this respect, we can say that our algorithms are never 100% bias-free. We're trying to do that. But they are much more and more neutral than human behavior. If you are traveling in many other countries, if you compare it internationally, does the law in Germany and Europe limit you? I think that the EUAEI, as it is now, and how he thinks that the European Union is good, because he creates a legal framework that is reliable and is also usable for us. I think that what is very important is then also in the implementation and in the national application not to let a flicking carpet arise, such as when we had data protection regulations that were reliable, so that you as a company can not only act uniquely in Germany, but ideally uniquely in Europe. But I think the big regulations as we have today, the EUAEI Act, the Digital Markets Act, the Service Act, are all good legislation that first create a legal framework in the digital space within Europe, which is unique and which can also be used by all countries uniquely. And that's what we do in Germany and in other countries. The Digital Services Act, there were a bit of turbulence in the negotiations with the US in principle, should be sacrificed that big US tech companies are allowed to participate in Germany. You certainly voted against some others massively. What is a demand that you would put on the politics, that is not to be given too much in order to come forward at other places in the cells? So now we are discussing the Digital Services Act, which is what I have made for myself, the discussion that we had before the Summerfane, the Digital Markets Act, where, in fact, which was also part of the negotiation in Zollstreit, the Digital Markets Act rules a fair competition in the digital space in Europe. And that does not mean anything other than that in the digital space it is exactly the right thing to do, even if it is in the economy. We have free markets. Everyone can freely try to speak freely, but we are not allowed to use the monopoly situation. We have always had that. In Germany we had the monopoly commission. In the digital space there are so-called gatekeepers, i.e. services, through which almost all Internet users have to use for the first time in order to get into the Internet where this competition exists. And the concern is that a gatekeeper, and to count on this, such as Google, Microsoft, Apple, that the gatekeeper uses their position to put in the foreground. And the Digital Markets Act rules that. That is why it is an important legal instrumentarium. That was also taken before the summer break again from the negotiating table. In this respect, I am very satisfied with it. How do you generally see the development of AI in Europe, Germany and Europe? It is often criticized that Europe in principle only has one own large language model that can participate internationally. That very little happens on the national level. Digital sovereignty. Do we need it in Europe? Do we need a own large language model? Or are there so many AI applications that everyone can already build their tools? Yes, that is a bit of a look in the glass ball. What will happen? How do these markets develop? It seems so to be that there are many competing large language models that also bring up fast technical prospects. My opinion and also our opinion is current. For example, if we take a commodity, it sounds weird for such an incredibly innovative application that is presented as a basis technology for a lot of applications that are built on it. I don't think that you can cut the discussions about artificial intelligence on large language models. There are many other applications that are also artificial intelligence that have nothing to do with large language models or build on large language models. Where we need sovereignty is the data sovereignty. It will only be done if I connect the intelligence with data, with business processes. We also have a huge chance in Germany. We have this very, very strong industry. Of course, we also have many strong national champions. It is important that we connect artificial intelligence algorithms with the data. So the company is not a requirement for politics, but for the economy. It is not easy to give these data out of hand, but to build your own AI application. As I said, the actual value comes from the data. The value does not come from large language models, but the fact that we can train them with millions of job decisions, with millions of jobs, with millions of lives. That's why we are able to really offer a good matchmaking. That can't be artificial intelligence, but you have to contribute to it and you have to contribute to it in an ethically correct way. There is now a economy that gives gas, that asks itself what kind of data I have, with which I can really build unique things. By the way, I encourage many companies that this is happening. We may need a bit of patience, but we also need a bit of patience when we are just watching. We definitely need patience when we are just changing it. But there I am good at things. What does that mean for the companies? For example, in the field of leadership. To go through the changed conditions through KI. Yes, there are two things that are very important. The first is, how do I actually lead through this transformation? Because maybe we are all excited about the possibilities that KI offers, but at the same time we have a lot of people afraid of artificial intelligence. And that is not a very abstract fear, but we have just talked about the work market data. Also a very specific fear, namely, will there still be my job in this form? And I believe that as feedback, we are fighting right now so that a latent enthusiasm and an abstract enthusiasm is there for it. But if it is about the specific application, KI is not used consistently enough. That is also no wonder, because no one likes disrupting his own workplace. So that means I need a good mix. Of course, I have to take my employees on the journey. And carry out training for them, so that they can use artificial intelligence. But I also have to make decisions. What are the processes, the business processes that I now digitalize? Where will I deal with more people in the future? Where less? And where maybe no more people? And these are the decisions that I have to make today. They are not always just nice and comfortable decisions, but they belong to uncomfortable decisions. Just like that in the industry, in the many decades to this day, it was also the case. So that it is a transformation of successful people. The second topic is, how do I not only lead people, but also the artificial intelligence? These algorithms basically work as employees. But as inexperienced employees. So I have to train them to do their job. I have to understand how they think. And I have to understand what chances there are. But I also have to understand where I should perhaps control, where I should push back. And to understand that an artificial intelligence that makes me more productive and that I have to invest time there, that is very important. In this case, to lead a hybrid, I lead a team of employees, but I also lead a team of algorithms. That is very important. In this case, I did not have to become a large-language model engineer, but I had to become an experienced user. I had to know what they could do and what they could not do. In this case, it will be a hybrid model of human and technical leadership. And is there already the right understanding of the company? Yes, in the USA, there is a current trend where the topic is started to be discussed. I would say that it still has this hype phase. So at the moment it is a cool topic. It is experimented. Then it will probably disappear in the classical hype cycle in the next months. It will be bad experiences. But in a very natural way, it will probably belong to the very normal tool boxes of leadership powers of all employees, but also to the leadership powers. So yes, as you can see, there are beginnings in Germany. In the beginning and in the surface, the topic is discussed, but not in depth. I think that we still have to learn a lot about AI. We also have to be brave to enter existing structures. Sebastian, thank you very much for being in the Table Podcast today. Thank you for the invitation.