The Agentic Enterprise

Transcript

00:00:00Welcome to ThinkDifferent, ThinkAI, the podcast from Mark and Jens.

00:00:07Two technology-loving minds who not only talk about artificial intelligence but live it.

00:00:14Here you will find clear classifications, real practical insights, and a fresh perspective on what is possible.

00:00:20Understandable, critical, and always with a wink.

00:00:24AI for reflection, for a smile, and above all for conversation.

00:00:35Another week has passed and we would like to warmly welcome you back to Sink AI Sink Different.

00:00:42Today, Jens and I are not alone again, and we are very happy to have Martin as our guest today.

00:00:51And with that, I’d like to start right away, before Jens tells us something we all know.

00:00:55We all know Martin. Who are you? What do you do? How do you engage with the topic of AI?

00:01:00Hello Mark, hello Jens. It’s an honor and a pleasure to participate in this podcast. Who I

00:01:07am, the short version, Martin Wurzmann, I was with Volkswagen for a long time, 9 years the

00:01:14group CIO, I learned there how large corporations function, how transformation works

00:01:21and also how it does not work.

00:01:23Then I faced my trauma for 2019

00:01:26and then I went to Silicon Valley for three years for CES.

00:01:31There, they totally threw a wrench in my plans with the move.

00:01:34But that was okay, it was a lot in home office.

00:01:37But there I really learned how tech companies think,

00:01:41how mindset can shape companies.

00:01:44Maybe we’ll revisit that later.

00:01:46The RTTU, the TÜDE.

00:01:47And then I was two years in a startup at Volta Trucks as CTO and learned what it means,

00:01:56to have to do a lot with little resources in an offline startup.

00:02:00And that was a great experience.

00:02:03Yes, and the topic of AI has accompanied me through all three stages.

00:02:07I founded the Data Lab at VW in Munich, in 2014, very early with AI applications

00:02:14in large corporations, in production, in sales.

00:02:17Yes, and in tech companies in California, AI is part of the DNA and in the Sada families with AI, we built the first organs,

00:02:26because we didn't have money for proper software, and then we thought of another way to solve the problem.

00:02:33And today you are our guest, which is wonderful.

00:02:35Just a quick hello also from the other corner of Germany.

00:02:38Hello Jens, I announced you just now.

00:02:40Nice to have you back today.

00:02:42I was a bit worried earlier that today might be cheap, but nice that

00:02:48you are attending today as well.

00:02:49Yes, welcome from my side too.

00:02:52Great to have you here, Martin.

00:02:53And of course I like.

00:02:54It's Christmas time, there are many other appointments, but of course we try

00:02:58to continue recording our podcast regularly.

00:03:00Because the topics don't run out that can be discussed.

00:03:04And Martin just brought up a few things and also

00:03:08for the podcast preparation, I stumbled upon one of those topics that

00:03:13always kind of triggers me, the speed at the moment and these

00:03:16old sayings, yes nothing is as fast as right now and so on

00:03:18and blah blah blah, but honestly, you have been in

00:03:22the information technology space for a few years now. What we are experiencing right now and also

00:03:27before we dive deeper into the other topics, I

00:03:30me myself, and I would also like to know how you experience this. I

00:03:34I find it almost burdensome how much unchanged is actually happening out there.

00:03:40I have this issue where I can't really absorb all the brilliant, amazing information

00:03:45that is happening right now, in order to process it properly and produce it

00:03:50and maybe also present it for others, so that one can actually do something with it.

00:03:53So I have this sense of responsibility, but I also notice that even though I

00:03:57am technically enthusiastic and have been involved for years, it’s really moving quite fast right now,

00:04:01and regardless of which news you look at, you just can't escape AI,

00:04:07not even in our bubbles that we listen to. But

00:04:10it’s pretty cool what’s happening right now. And I don’t know how you perceive it,

00:04:15but I have a few thoughts on that.

00:04:19People around me, to then shoot the bow. I have a few people around me,

00:04:22who almost feel overwhelmed and then almost have blinders on again

00:04:26attach and say, stay away from this stuff for now, it might also go away in that moment. I think you made that point quite well. So,

00:04:29it might pass in that moment. I think you made that point quite well. So,

00:04:34I always compare it this way, until a year ago we were used to in IT, innovation cycles,

00:04:42but the software had two releases a year. There was the Spring Release and the Fall Release

00:04:46or Summer Release, which is now SAP or Salesforce. You always had two releases

00:04:51and you knew there would be 20 new features, twice a year. For a year now

00:04:56it has been broken down to today, every five days. I just read that,

00:05:03Major Releases, we are not talking about minor things, we are talking about groundbreaking performance improvements,

00:05:08new features coming from OpenAI. And hardly has OpenAI released it the next day, when

00:05:14if it was done skillfully, then Google or Entropic or Microsoft or whoever comes around the corner.

00:05:20And that's why it's so quick, I listen to several podcasts and notice how everyone has the same problem, how you just want to keep up with the fast, short headlines. That’s definitely a challenge.

00:05:36I think what changes is that it’s not just a pure IT topic this time.

00:05:42In many cases, looking back into the past, new browser technologies or something like that emerged,

00:05:46new devices, which naturally changed access, a cloud service came along.

00:05:50All these things fundamentally changed the IT topic.

00:05:54But this time, it’s an IT topic that can also be defined by the neural networks,

00:06:00that we now have and the capabilities that artificial intelligence brings.

00:06:03but it is of course groundbreaking for all possible areas, whether it’s now

00:06:07and I have talked about it again, whether it’s medicine or even if

00:06:10you now say, the various models that are actually conducting real

00:06:14scientific work. The research results are already being published

00:06:17and topics are being presented, where we as humans again do not understand what

00:06:20is actually behind it. I believe that is also what makes it

00:06:23so groundbreaking and extraordinary right now and where one has to somehow

00:06:27see that one stays involved, definitely, as a person, as

00:06:30a society, as a company. And of course, everyone feels initially affected, meaning

00:06:36if I don't understand it, I'm still affected and I feel that something

00:06:42is happening, probably out of fear when you don't understand. And when

00:06:48I do understand it, I notice even more and ask myself what that means for me

00:06:52personally, for my job, for my everyday life, for my daughter who is 21

00:06:57and currently studying, you constantly ask yourself those questions. We didn’t have that two summers ago.

00:07:03No, I also always think, when you are busy with all this stuff and I always have my

00:07:09Sunday cycle, meaning throughout the week you engage a lot with it,

00:07:14because you deal with it in the office, because you deal with colleagues with it,

00:07:17then comes Saturday, where you catch up on all that, which has somewhat been left at home during the week

00:07:21and on Sunday you sit there and think

00:07:25what the hell, while some might be considering what this beautiful term is,

00:07:30what kind of technology is coming towards us, like, I don’t know, things that you

00:07:34probably wouldn’t hear arrogantly, despite having dealt with it some six months ago.

00:07:37And you think, what the hell, this pilgrimage, we’ve already come further, we

00:07:41see quite different hurdles and challenges and terms on the horizon, where you sometimes

00:07:47start to ask yourself how to take the people, the people along

00:07:52on this journey, and how to continue on this journey in a way that you, yes, I’ll say

00:07:58now, do not get left behind. I mean, we have also made it our task in the podcast

00:08:02to explain a bit what we are actually talking about. Today we wanted to

00:08:06talk a bit about how agent AI, what transformations it brings along.

00:08:11Would you briefly tell us what you understand by

00:08:16agent AI? I will try not to get into too much detail

00:08:21because you could go on for hours about that.

00:08:23The laps organizations are delayed by an hour.

00:08:26Exactly, so if you take the topic of AI and sort of

00:08:30cut it into the developments,

00:08:33it all started at some point with simple statistics,

00:08:36data evaluation, value Data Analytics,

00:08:39English term.

00:08:40Then came this, which is by the way to explain,

00:08:43statistically, what data has to say.

00:08:45The second step was then Sheen Learning,

00:08:48where it already went into forecasting.

00:08:50To say beforehand, what could happen? Those were the beginnings of AI

00:08:55a few years ago, when companies also started to massively

00:08:59invest. That is the second step. Then at some point the third step came,

00:09:03the LLM models, ChatGPT, the first version, the chatbots, where it is about recognizing

00:09:10language, breaking down language, understanding language and then creating new content

00:09:15translating it back into language in order to

00:09:18give it as a response. Isachwas, just a brief answer. These are the chatbots, ChatGPT. The next evolutionary leap

00:09:26was then a step after that. Agendic AI means I have the LLM model, but I can communicate

00:09:34with an algorithm. Isachwas and get something back. But the second step is that

00:09:39this model can access all kinds of data. MCG, this Model Context Protocol,

00:09:46technology, technology, where suddenly I can tap into other data sources.

00:09:51The third point is that this model executes, meaning it does not just give me an answer

00:09:59and says, the cheapest flight from Frankfurt to Zurich is the one with Swiss, but I

00:10:05book the trip right away, meaning it is executing. And the last step is the

00:10:11so-called learning. This means, the algorithm, the agent,

00:10:15updates itself in a database about what it just did. The next

00:10:20time it does the same again, it looks at what it did last time and sees if

00:10:24it can improve. So this way the agent learns and, let's say, connects

00:10:30off from any standard we have set. It becomes

00:10:34independent. And I think that's important to understand. A chatbot, too

00:10:39like our call sender, is not an agent that answers a question, it simply

00:10:44accesses information. It doesn't execute anything, it usually doesn’t do anything for me, and

00:10:48it doesn’t learn independently. So if you think about it, one is

00:10:53certainly faced with, I would say, greater challenges.

00:10:57We touched on IT a bit earlier, and there it’s also about the

00:11:01releases that come to market. That seems to initially throw everything

00:11:05upside down. It's a bit like IT; I myself am from IT, by the way, so

00:11:10greetings to René, in case you hear this—some of you know him—he has been a guest with us

00:11:15already. Let's just say the IT has to somehow reinvent itself here. So it probably has to

00:11:20set itself up differently. Do

00:11:26you have an assessment of how this will change, how it should change?

00:11:30Yes, I found, perhaps with my perception—and maybe you have a different one—but

00:11:36one would assume that especially IT people would immediately jump on such a new topic,

00:11:41like a bull onto a cow.

00:11:42They exist.

00:11:43But you also find just as many who are extremely skeptical.

00:11:48And where do we come from?

00:11:50We IT people always say we come from the ERP era, that is the Stone Age.

00:11:56Back then, processes were cemented, glued into an SAP or another

00:12:02system, to ensure processes run stably.

00:12:04That was also the right thing, from the 90s to the 2000s, and you even modeled in

00:12:09Arles, that it's really glued down.

00:12:13And now there's a change in the business, for God's sake.

00:12:17Any change is extremely expensive, time-consuming, and you risk that your clean system

00:12:22suddenly doesn't work anymore.

00:12:23It costs us a lot of money, takes time.

00:12:25That's why we have IT people leading change management,

00:12:28to prevent changes.

00:12:30Change management is nothing more than

00:12:32a long process with committees to prevent

00:12:35anything from changing.

00:12:37And we know that exactly, and they also don't want

00:12:40any of the established processes to change.

00:12:43So we fragmented it, IT with the business,

00:12:46to change nothing. That's the IP world.

00:12:49Now you suddenly have a technology,

00:12:51that is completely automated and has partly developed on its own.

00:12:57So an agent that adapts, you can give it, it can be a process modeler.

00:13:02So, we find it a bit difficult coming from the history.

00:13:06Yes, I think that's an exciting point because I also mean this, if you say,

00:13:11an agent that adapts and also learns on its own.

00:13:13So, that's also a topic because I've just read briefly now,

00:13:17Google has released a paper somewhere, where it was about Nested Learning, where

00:13:21once again another way in which agents then afterwards also have knowledge, parts forgotten.

00:13:27So it's really about two layers of memories. So things that are stored for a long time

00:13:34in the AKI, things that are rather stored short-term and need to set in more.

00:13:39So again, quite similar to human behavior, how we learn ourselves.

00:13:43And those are all things. I say that's IT in the background, but it feels 0 like IT.

00:13:50IT and it's also really with the IT rings 1.0, if you want to put it that way,

00:13:54honestly not really tacklable, because it has something completely different to do. It's like, we've

00:13:59often talked about here, you know, the people who are outside know

00:14:03the topic, when I prompt the AI to behave like a

00:14:07professor, like a consultant or some other stories, and that you get

00:14:10certainly different answers. And all of this plays into it. And that's,

00:14:14I believe that’s something, now I’m just a sort of half-IT person, where I say, yes,

00:14:18it’s also something that I, of course, see when I look at the IT world, that even before

00:14:22such a classic Data Scientist person, who got involved with AI early on, which you just

00:14:27described, those dealing with AI before this LNM moment have partially the

00:14:32biggest problem with solving and thinking differently at that moment,

00:14:35because from their statistics, but in the background from their mathematical background simply

00:14:40they are so used to that exact results come out at some point, which

00:14:44you can no longer expect in this world. And I believe

00:14:47we need to completely redefine IT, and business of course too.

00:14:50Yes, I would like to mention the example of flight schedules again.

00:14:53When the tradition of IT flight schedules means, I go to the website,

00:14:58I then enter my flight date, my destination, everything,

00:15:01click, look for the options, then choose the flight I want,

00:15:07then I go into the payment process,

00:15:08enter my credit card, and finalize the booking.

00:15:11This means the process is predefined,

00:15:13then I specify in the SAP system or on the websites, it doesn’t matter.

00:15:18Agent functions work completely different; I specify an outcome, a goal.

00:15:22I say, let’s find the cheapest flight.

00:15:24How you do that doesn’t matter to me at all.

00:15:26You can look at 15 websites, or just one, or book immediately.

00:15:31You don’t dictate the process.

00:15:33And we come from a world where we are very process-oriented, that’s how IT works.

00:15:38And breaking this way of thinking is incredibly difficult and uncomfortable.

00:15:43when you are faced with it, because you can no longer evaluate it based on your experience.

00:15:47That’s why many struggle. At the same time, I also notice there is a

00:15:53new kind of curiosity, because we are all playing around with VAT people at home with JETSCHIPETI,

00:15:57when we come to the office in the morning, everyone automatically thinks, how can I somehow use this

00:16:02for myself. That thought is somehow present for everyone. A new kind of curiosity market,

00:16:07these parlor names in recent years, that’s correct.

00:16:10It’s a bit like this, well, I’ll put it this way, people who have certainly been successfully doing their jobs for many

00:16:17years have also been shaped by

00:16:23the processes and the environment they are in.

00:16:26This means this attitude and this mindset, with which you, let’s say, approach the

00:16:31Work you engage in, with which you interest yourself in new topics,

00:16:36will have to change, because these frozen processes,

00:16:42no, frozen processes is wrong, this stone hewn, let's put it that way,

00:16:44they will no longer exist. Yes, it is also this, so I believe it’s not only

00:16:52related to the IAT. I believe, as a society, we are so used to the fact that we, our

00:16:55entire learning is actually determined so that we, that we sometimes take this,

00:17:00this path. You just said it beautifully, Martin, that one actually is,

00:17:03Now I can just give a magician a task, and he solves this task without me having to walk the path myself.

00:17:09I don't have to conduct the flight search now.

00:17:11I have to go through the booking process.

00:17:12I don’t have to remember where I just put my criticism to enter the security number.

00:17:16All these things I can outsource, and thus the path is no longer the goal.

00:17:20If I now think back to my multimedia training back then, where I would have programmed the first CD-ROMs and websites.

00:17:28And we already had the first visual editors that were on the horizon back then,

00:17:33which still produced disastrous code.

00:17:35And of course it was good that if you also used HTML or

00:17:38any other simple programming languages, you could do it.

00:17:42That you could at least enter it in a text editor back then.

00:17:44That was still such a nerd principle.

00:17:46Yes, and I don’t even know if that still applies today.

00:17:48You know, this fundamental question of how much one must still understand topics when in the background

00:17:54there is a machine that can actually do everything.

00:17:56So how do we as a society, as humans, need to adapt, so that we don't have to fully understand everything down to the ground anymore.

00:18:06I believe, I mentioned this earlier in another example, that we now receive scientific results from AIs, where we scientists wonder how they came about.

00:18:13What physical properties did the AI use to recalculate some magnetic field or something else?

00:18:20And that's simply not clear anymore.

00:18:22I'm asking myself the question, do we need to be a hundred percent close to all of this, or do we need to learn a new kind of trust towards these systems?

00:18:30Of course, with all the things, it should still switch to green behind a traffic light at the right moment, and if not, it should switch to red or something else.

00:18:39That is indeed a question we need to ask ourselves and which we need to approach with curiosity; I think that was a nice sentence from you just now.

00:18:45that we can rediscover such curiosity to rediscover the mirror-piss,

00:18:48where we really need to carefully consider how we set ourselves up. And here, IT comes back,

00:18:52in my opinion, as a player, now from my design business perspective in such secure

00:18:56playgrounds, even in a corporate environment. How can we actually achieve such things?

00:19:01I am currently writing in the book and one of my chapters, I love it too,

00:19:06the Ancient Factory. And that is a concept. The idea is simply to use this curiosity to

00:19:12take employees and assign them the task of designing the agents themselves.

00:19:17How do I do that? It's a studio. It's a small space, a virtual space.

00:19:24And in there are IT people along with the business people, and I also say to include HR people

00:19:29as well. And a cross-functional team. And they build the agents themselves. And actually

00:19:35in the first step, they describe what the agent is supposed to achieve. It should always book the cheapest

00:19:42flight, booking in a normal company is naturally different, but you

00:19:46describe the goal and this goal defines what it needs to do that,

00:19:51right? So you take the curiosity, they can also experiment,

00:19:56entering this protected space and the Boundiagetten themselves like Hausmacher. How

00:20:02Hausmacher, liverwurst in the Palatinate. I come from the Palatinate, everything you

00:20:06what you do yourself is always better. So this is the fundamental mindset here and it

00:20:10is the same there. And you don't need consulting firms and system integrators,

00:20:14who have their own people doing this, yes with a certain amount of guidance and support.

00:20:18And then fears are alleviated and creativity is utilized. And I have

00:20:22observed this a few times in companies, a genuine eagerness to experiment with curiosity.

00:20:28And when the first agent they built goes live in a small environment,

00:20:32the faces of pride and enjoyment, it's contagious.

00:20:37And it is finally the chance we've been waiting for, I believe, also in IT,

00:20:42that we can bring this closer to people by allowing them to experiment,

00:20:47if it is not that complicated. Then let me ask a follow-up question.

00:20:53When you say this about experimenting, is it maybe also now,

00:20:56the one or the other transformation I had the chance to accompany,

00:20:59is it now maybe the time that we can no longer separate IT

00:21:04and business in the future, so. Maybe not today,

00:21:06but in five or six years, can we still really distinguish it?

00:21:10So to put it bluntly, if somehow the HR manager you just described,

00:21:14can actually set up an agent system that he would otherwise have purchased,

00:21:18or that he would have had to deploy

00:21:21or have implemented by the IT department.

00:21:24Where is the boundary actually?

00:21:26So are we losing a bit of this role clarity?

00:21:31Is there still IT-IT in that sense?

00:21:33or is it rather reduced to something like it perhaps used to be,

00:21:38you know, setting up servers, the classic administration business at that moment

00:21:42and the actual software business basically slips into the hands of everyone.

00:21:46That’s just a good question you have. So I, I, I wouldn’t put it that way,

00:21:50I, so I, I believe I know where you’re going with this and I suspect that too, yes.

00:21:54This separation never really drove me before. We had it because the organizations

00:21:59want people that you can assign to a box. But we already had it in the past.

00:22:03You have many people in IT who understand the business better than their own business.

00:22:07And you have people in the business who understand IT better than some IT people. But that

00:22:12is simply the case now. Anyone can program something without programming knowledge. And I bet

00:22:18in a year, we will laugh about this conversation because there will be something else that is

00:22:22unimaginable today when every five days, as

00:22:26said, a new major release comes out. So I believe it’s merging. I also believe,

00:22:31this is the chance for cross-functional teams. So an ancient studio, design studio, where

00:22:38people sit together, yes. It doesn’t matter whether one comes from IT and the other

00:22:42from sales and the third from HR, they are building new digital employees

00:22:48together. As you say, then the background will eventually be irrelevant. You need

00:22:55of course, there are certain hardcore IT functions that you need,

00:22:58IT security, security, you have infrastructure issues, you don’t need to

00:23:02bother a sales person with those, you will always have them. And you have

00:23:06hardcore tasks in marketing and sales, for example, you also won’t want to

00:23:11bother everyone else with that. But probably 80 percent will

00:23:16overlap in terms of capabilities. On the edges you have

00:23:21but only the hardcore experts.

00:23:23I found it quite interesting that you mentioned those colleagues,

00:23:27colleagues are basically making their digital employees or digital colleagues.

00:23:31Do you see, the colleagues are basically creating their colleagues?

00:23:34Or how does it look with scaling?

00:23:36Or with the reuse in other areas?

00:23:39I mean, when someone says, I do the requirements engineering,

00:23:42you find that in many different places.

00:23:44Are these then things that you, I'll say, if they have reached a certain

00:23:46quality standard, can be generalized to the

00:23:49or do you think it will lead to all teams in the future, I’ll say,

00:23:54locally optimizing in a pristine world, where all the data is clean and who knows what else.

00:24:00How do you see the boundary there?

00:24:02That's a good question. So I think, in the beginning, it’s important that

00:24:06the more you dive into such a topic, the better.

00:24:11I wouldn’t want to regulate everything right at the start, nor set standards.

00:24:16We love doing that.

00:24:18We always like to define in Scharnfeld first, then we get started.

00:24:20But the point will come where it will probably get chaotic.

00:24:24Where you will also lose synergies.

00:24:26The question is, where are the real turning points.

00:24:29I know, the platform manufacturers of, so the Agentic platform,

00:24:33The topic of H-Sheet Orchestration is a huge topic.

00:24:38Firstly from a governance perspective, but also about sustainability.

00:24:43So that it isn't developed five times in the clear agent, it would be a waste.

00:24:47But I think this is an evolutionary process.

00:24:49In the beginning, I think it’s important for people to just get started, do it, just do it.

00:24:54And then one needs to develop a feeling for the right time,

00:24:58to bring it somewhere into a track.

00:25:05So, this, what you meant by a desire for new things, a desire to experiment.

00:25:09That one gives people the space, on one hand.

00:25:13It's always nice to be able to try something out,

00:25:15without immediately getting your head bitten off.

00:25:17Here you can try not to engage with the new technology,

00:25:21but on the other hand, of course, have an insight into it.

00:25:25I mean, not everyone is a corporation.

00:25:27We are all somehow connected to corporations,

00:25:29we at least have points of contact,

00:25:31where governance and so on are written large.

00:25:34But simply creating that safe space,

00:25:36so that people can try things out”,

00:25:37and when good ideas arise,

00:25:39to create the right environment for that,

00:25:41so that you can act as a driving force

00:25:44and can broaden your scope. Yes, the exciting thing will be, as I listen to you right now. I believe,

00:25:50the game is important, definitely no question about it, we have to do that. I mean, that's also a

00:25:54reason, Mark and I also play podcast to engage with the topic as well. That

00:26:00is also a kind of game that we play to delve deeper into these topics,

00:26:03to have intelligent conversations, to approach them. The question is always like

00:26:09of course from a corporate perspective, from a business perspective. When is that moment

00:26:15reached? When does something have to come out of the playground? And how certain, because I found

00:26:20it nice that you just mentioned the topic of chaos, how certain can we actually

00:26:23be that this wild, we just heard it in science now,

00:26:28that we don't really understand it, the black box. The black box gets even crazier

00:26:33to be honest, when we say we not only have an AI that we don't understand,

00:26:35but we have this AI in such an agentic system, in such a black box system.

00:26:39connected with each other, can solve various tasks,

00:26:42because then we also have part-processes and tasks in the playground

00:26:46that we have let the AI handle almost fully automated or automated.

00:26:51There will always be that moment when I say,

00:26:52now it has started productively.

00:26:54And in the productive systems, I don't have just one AI,

00:26:59that I restrict with a few guardrails or something,

00:27:01but I really have an Agentix system out there,

00:27:03that takes over processes externally, takes in processes internally, how strong is

00:27:10the human in Belup still really keeping up, really, really in Belup, or how strongly

00:27:17does AI actually have to control that already? How much do you actually have to, when you do something like that

00:27:20we try from the very beginning, whether we also have AIs in this agency system

00:27:27as supervisors directly, because we may have to admit as humans

00:27:30that we can only be a little bit confident when it comes to human

00:27:35involvement in the profession and can actually control things, honestly. So that's a

00:27:38bit of what I always think when I look at these things, where I say, when we,

00:27:42we've been talking for three years about salting, whether that's good or not,

00:27:45all of that has been put on hold. But then you have these agency networks,

00:27:49where it's really going to matter that multiple AIs will take on the tasks

00:27:53and execute them, and maybe also old deterministic systems at

00:27:57one point will produce very clear results, and then you let that loose and it will happen multiple times in the

00:28:02world, then suddenly the flight website is no longer

00:28:06a simple website but also an agentic network that essentially interacts with our agentic

00:28:10network or your private one and that is a chaotic system and I don't believe

00:28:16that we humans will be able to control this system and I'm saying this

00:28:22not out of concern but rather out of excitement I think we need to accept that

00:28:27we also have to entrust parts of this chain in a trustworthy manner to the

00:28:33topic of AI because I don't believe, maybe you see it differently, but

00:28:36I don't believe that we can fully control it, you both are more IT people than I am

00:28:39and feel free to enlighten me with something better, but I believe this will be a task

00:28:44that will honestly overwhelm us humans.

00:28:46Well, I have a different opinion, we should trust. Trust is good,

00:28:51is better, with agents not with humans. And you shouldn't monitor humans

00:28:59in every step they take. The Works Constitution Act has stress points with the works council,

00:29:04if you have a large labor representative in the works council, rightly so. Agents have

00:29:08no emotions and they need to be controlled rigorously. How do you do that?

00:29:14For example, there is an Agendic Journal. It means the agent permanently writes

00:29:18in a database every step they take.

00:29:21Every, every detailed step.

00:29:23So what prompt did I have, at what point in time, what prompt was it, which

00:29:28data did I extract from which source, and very importantly, every agent has an outcome

00:29:33to achieve.

00:29:34An agent works towards the outcome.

00:29:36Objectively, it can be measured, did he succeed, did he find the cheapest flight with

00:29:41Swiss from Hanover to Zurich.

00:29:43If there was a cheaper price, it's registered in the database, the Delta outcome becomes reality.

00:29:49That means, then the prompt is adjusted automatically.

00:29:52Then he will have to tighten the prompt to achieve it next time.

00:29:56And in this journal database, you can put another agent on it, who does nothing else but observe.

00:30:03How much is he deviating?

00:30:05If it’s too much, at a wide range, the station, if that’s 1000 for the flight, is it a mistake

00:30:12or is that really too expensive?

00:30:14So you can implement control mechanisms and that's why I say, you need to implement control mechanisms

00:30:19if they are very strict.

00:30:22And kills will be activated if the agent drifts.

00:30:25Then he will be corrected and will be rectified.

00:30:28And the responsible low-vote war is displayed on the screen, the agent has been stopped,

00:30:33Stop-comment-vote, then the intervention can happen.

00:30:35There's another term that I also don't like, it's Human in the Loop.

00:30:40That somehow makes us into lackeys.

00:30:42Expert in the Loop is actually much better.

00:30:44I didn't invent that term, I heard it a few days ago.

00:30:47It wasn't really good, yes.

00:30:49The human becomes the expert, who assesses the quality of the digital colleague workers

00:30:56and not as a guardian, but really, how can I improve this?

00:31:00So why did he react that way now and how can it be improved?

00:31:03And promptly possibly or maybe different data,

00:31:07maybe the task is set up incorrectly as well.

00:31:09This means you will be much more challenged in this expert role.

00:31:14He just sees it as a little snooper,

00:31:16so that he can be part of it and checks if everything is alright

00:31:19and presses the button.

00:31:21That is to be chosen.

00:31:22Actually a cool term, thanks to the one who mentioned it.

00:31:27I’m using it now too.

00:31:29And yes, that was the topic.

00:31:32I found it also very understandable, so what you just said,

00:31:35with Human in the Loop, that’s known as a term.

00:31:37That we are in the expert role, I think is cooler too.

00:31:40We’re noticing that too, yes.

00:31:41I mean, it’s not like we’re not also dealing with a few

00:31:44agentic players and so on,

00:31:47where we then say, even this measurement is what he does, the right thing.

00:31:51Because there are also so many other influencing factors,

00:31:53not that he perhaps didn’t find the cheapest flight,

00:31:56but then comes tomorrow again, whether more eggs around the corner.

00:31:59throws a new model on the market and suddenly you have to watch your prompt, making sure it still responds and doesn't just revert to what you previously tried to train it out of.

00:32:09He’s now somehow executing that even more excessively due to his new model.

00:32:14Yes, that’s basically a constant, calling it supervision sounds harsh, but for the machine, that’s allowed.

00:32:22We had this topic that the digital employee can indeed be very monitorable at that point.

00:32:29We were just talking about companies, so if we look again at a larger

00:32:35organization, what do you think, how will agentic AI, let's say also in the context of

00:32:41a company, manifest itself in the coming years? I mean, who has been focused on

00:32:48employees or on someone who, to stick with the example, finds the cheapest flight

00:32:52available. But how do you see the whole topic of agency for leaders?

00:32:57There is a company in Switzerland, the market leader, retailer for mobile phones and mobiles

00:33:04with mobile contracts. About Switzerland, you have the small jobs and they told me

00:33:09that the CDI-Opt told me, they introduced agents in the Arctic departments.

00:33:16So teams, there is a person inside and agents, they have a name, they also have a personnel number.

00:33:22Microsoft has done that too, I read once.

00:33:25But in the team meetings, when everyone is on the team,

00:33:29is an image of an agent, who once picked out such a picture, along with it.

00:33:32So they also humanize the agent.

00:33:34And he is included, he has a clear task as well.

00:33:37And it is clear, certain tasks are now performed by the agent in the team.

00:33:44In the past, the employees did that and then it is essentially integrated.

00:33:50I found it quite interesting, he mentioned several aspects, how to address fears,

00:33:54how to bring agents and humans together in an environment that somehow feels human.

00:33:59He told me that the agents are present during the team meetings,

00:34:02when we go for a beer in the evening, of course not, but that was not really, really good.

00:34:09and they experiment with how employees and the organization react to such

00:34:14organizational forms, such hybrid forms.

00:34:17Let's not kid ourselves that we will be in the future where more and more of these

00:34:21agents take on some task in your team and probably

00:34:26there will be fewer and fewer humans, yes with the demographic curve, which is declining.

00:34:30We will have fewer employees in the next 20 years, we will see a decline

00:34:35due to retirements, it will automatically happen that you will always replace them or perhaps not even want to anymore,

00:34:42to save costs. And then we need to deal with this issue massively.

00:34:47When we deal massively and how the whole topic of AI agents is mentioned repeatedly,

00:34:57how, I would say, enterprise-ready it is regarding the technology.

00:35:02So, you just heard that there are already companies that I would say, let AI work alongside the staff.

00:35:11How operational is it, do you see the whole thing? Can we now say, okay, go ahead, or maybe pay attention to this and that before you start?

00:35:21How operational do you consider it?

00:35:23Well, I would say the excuse that the technology can't do that is no longer valid. Luckily or unfortunately?

00:35:30Yes, I'll go back ten years, integrating five legacy systems, databases,

00:35:35you cross.

00:35:36Yes, and partly the data didn't match, the formats didn't match, that

00:35:40was, there was middleware at some point, it was easier.

00:35:42Today, today I integrate personally and we proceed together when we sit down

00:35:48in front of screens.

00:35:49I won't name any product names, so off-the-shelf products.

00:35:52With three clicks you have SAP S4, every table you want in any field.

00:35:59integrated and pulling the data out and matching it against data from Salesforce or

00:36:06something else. So the data integration time, which was so difficult, is today, I would say,

00:36:13no longer an issue. I also do not accept the argument that the data is so bad. I believe

00:36:17No, then the company won't function at all if it were that bad.

00:36:20We're high up and maybe not as prepared as one would like and need.

00:36:25It will be like that, that's what Tunis is for.

00:36:28Then these LLM models are good enough to have simple OZ steps as a goal, results.

00:36:37That is definitely feasible.

00:36:39Whether you want to develop the next Golf with agents, I doubt that today.

00:36:43Yes, but handling the procurement of indirect materials with agents, absolutely sure.

00:36:49Monitoring the supply chain with agents who then reallocate materials.

00:36:54if there's a problem somewhere, if trucks break down, absolutely, yes. You can build that today.

00:37:01So there are use cases that are absolutely feasible with what we have. And a few others,

00:37:07you don't necessarily have to tackle on the first day. You might be able to put them on hold.

00:37:12pushing the pregnancy. It's also a bit about bringing successes to the people

00:37:17that when you've implemented something, then, I would say,

00:37:22look, it works. Yes, because you just mentioned earlier

00:37:25about migrating databases and so on. I immediately think of people

00:37:30who have accompanied those things and who now look at you with question marks

00:37:35in their eyes and say, that's what the AI should do in the future. I still remember,

00:37:40when I walked into the office this morning and shared my findings from the weekend

00:37:43about how here, when one has not only said, look at

00:37:47nice pictures, but you pose a question for, I would like this to

00:37:52offer this function, that must be articulated in the end, it should

00:37:56connect, no idea, another three sentences and two, three, four, or even ten

00:38:01minutes later you have a first prototype in hand with source code,

00:38:05on which you work, and then you slowly move closer to the

00:38:08next solution, where in the past people would have certainly rightfully

00:38:13spent longer on it and now sit there and say, well, okay, I will bring my

00:38:18expertise in to give this piece of code that pops out

00:38:22maybe the next shift, yes, because I say there's

00:38:27also a difference whether you, no idea, I told him

00:38:30recently to build me a teleprompter application and when

00:38:33there are questions from the audience, it should also display the answers

00:38:37right away. That didn't happen, he just built me a teleprompter

00:38:41and I thought, how cool is that, but surely there might still be a

00:38:44small difference to me doing the large cm-system for whatever.

00:38:48I don't know everything but a little bit yes, although I'm a bit of a

00:38:53exactly half of many here say in the round yes, I'm a bit of a

00:38:59I don't even know, I remember hearing a sentence from you in this

00:39:03direction, I believe we are sometimes too fearful. I think especially what

00:39:07you are saying, Mark, is such a classic way of thinking that, I believe, naturally you are brave

00:39:13in the applications you are involved in, and you also drive us forward in many ways.

00:39:17But I really think it’s not that, but a CRM system, I wouldn't quite

00:39:22let AI handle that yet. Why not? So why not? So really, I've

00:39:26already seen things you've coded, briefly prompted, where you really have partial functionalities

00:39:34in a perfection where you say, there essentially is the authorization process,

00:39:38I could log into this application with Google, Microsoft, whatever, that is

00:39:42everything had already been arranged, and then there was an image upload, connection to

00:39:46Google Calendar, of course, a small thing was built, and it all works first. Why

00:39:50shouldn’t I say nowadays, I also build the big application that I

00:39:56need, completely with AI, why not give it a try and maybe

00:39:59the Martins Playgrounds idea? And then there’s the lab here, the whole grain independently

00:40:05the right specialist departments. And then they should see if the big software solution,

00:40:10which was otherwise purchased expensively or took three years before, can migrate to here.

00:40:13Exactly. Maybe not in a week.

00:40:15Exactly. Just an example. What does a CRM mean? One needs to break down the more complex

00:40:19picture. For example, they don’t need everything at once. Maybe you say,

00:40:23I need every phone call, every email, every contact in essence at the beginning,

00:40:27I don't want to type that in, the CRM should be automated. Such an agent

00:40:32you build it with access to the email, with access to the transcription of the call,

00:40:38everything is already possible today, and you can build it in a nice evening with a glass of wine,

00:40:42then you have this feature. This is a CRM feature, automating customer outreach.

00:40:48Then there might be the next thing, creating messages from that. Customer letters,

00:40:52like those who haven't reacted in six months.

00:40:54So, with the goal, mainly focusing on the outcome, what should happen,

00:41:01the outcome, then you can build the agent around that.

00:41:05A CIM or SAP system doesn't have an outcome as a goal.

00:41:10The goal is to have all the data together, to have all processes mapped out.

00:41:15That is the goal of a large system, if any other is counted,

00:41:19When producing results.

00:41:21And then it also takes...

00:41:24Maybe that's also the perspective that one might need to learn anew.

00:41:29We don't build whole systems, but just like we used to solve problems as humans.

00:41:35There was the carpenter who specialized in that topic.

00:41:39And then someone still had to get the utility installer,

00:41:42who connected the service and the planner, who planned the village.

00:41:45It's rather this interplay of these individual outcome-oriented functions.

00:41:52No longer, I need a database and then have this other layer above it,

00:41:56which contains the function calls or something else

00:42:00and then I have to tap into this and that library,

00:42:02and therefore I have to, because I have to do all of this,

00:42:04I obviously have to plan the energy first,

00:42:06so that it's ready in two months.

00:42:09I think this is a new way of thinking that I need to enforce.

00:42:12We no longer think in programming, in actual programming, but rather

00:42:17in soft various outcomes that need to come together to perhaps achieve a larger goal.

00:42:23For example, I have experienced myself that in writing a book, the biggest

00:42:29time consumer was the research, constantly searching the entire internet every day.

00:42:36Someone writes something on the same topic and publishes it,

00:42:41some kid or whatever they are called. It's so time-consuming that you could, as you

00:42:45know, easily spend 12 hours a day and still feel like,

00:42:49I've only caught half of it, if even that. Because

00:42:52my goal was, the goal was to automate this research in such a way that I

00:42:58receive a list every morning with the 10 top findings that concern my book

00:43:04and a suggestion on how I should respond to them in my book. The goal was,

00:43:10to have the book six months ahead of what it currently does. Because I know that the moment I

00:43:15stop writing and publish, it becomes older every day. And eventually, it will be outdated,

00:43:21probably. And then I built agents here. I created four agents that perform this

00:43:26research, summarize it, benchmark against my book, read my book constantly,

00:43:32where they read status updates and make suggestions on what I should adjust, and the outcome was,

00:43:39ensure that this book remains current in its quality and then you build

00:43:45exactly around that.

00:43:46If I had somehow built a book writing agent, I would still be transiting today,

00:43:51I really played around and experimented for 14 days until I had it

00:43:55figured out.

00:43:56And they even learn, I also tried this Continuous Dynamic Aging Pumping,

00:44:01they constantly improve their prompt. When they realize they are hitting the mark. And if the results

00:44:07aren't better than the day before, then the agenda and prompt get rewritten. That's really

00:44:12cool. I have no idea what they are doing anymore, but I get my daily

00:44:18list of ten points every morning for breakfast and I enjoy it,

00:44:22that someone is doing it.

00:44:23You just mentioned research, and when the book is published

00:44:28it basically races against time, because what has been put down on paper or as an e-book

00:44:33becomes outdated by its very nature, and I'd say in the agentic

00:44:38world that might happen even a bit faster. What's the book called? When is

00:44:41the book coming out? What is the book about? I mean, tell me a few more

00:44:45sentences about it, because, yes, I was allowed to see the cover at a presentation

00:44:49once, and I'm also interested in it from that perspective.

00:44:51It's called The Agentic Enterprise and the title is Building Organizations that think, learn and act autonomously.

00:44:59But it’s about the autonomous company that operates fully agentically.

00:45:04I'm building a story, jumping a few years into the future.

00:45:08If you push this further and companies operate fully agentically, what does that actually mean?

00:45:13And I have developed a framework, the Agentic Enterprise Framework, which has 12 modules.

00:45:18It explains how can I technically achieve this with what is possible today.

00:45:24So I'm writing it, it’s a fiction story, but it's based on, what's technically available today,

00:45:29how do I get there?

00:45:31Both, how do I build the organization and how do I take the people along?

00:45:34The entire transformation process is actually more about the people than about the

00:45:40agents themselves.

00:45:41How do I do that?

00:45:42How am I the one, the agent factory for example?

00:45:46The third is, they must react to leadership.

00:45:48What is the task of the CBO, who used to enjoy IT topics?

00:45:53I mean, yours is a bit different, but I know most CBOs who are supposed to lean back, let’s just handle IT.

00:46:00It costs a ton of money anyway, takes forever, so you handle it.

00:46:03We had a different role in the game.

00:46:05The second is the CBIO.

00:46:08Some do this on the side, bravo Röhrsmann, the man, and many others learn.

00:46:14But this is a huge task to navigate and lead in the new world, to guide IT.

00:46:21And the third one who is always forgotten is the HR chief, the CHRO in English, the HR chief,

00:46:27who is responsible for adapting the culture, to take, by the way, themes.

00:46:32Goals change, people have different tasks, fears arise, we will prevent blockades.

00:46:39The HR chief has such a central role in the transformation, more than ever, and that

00:46:44are the core elements in the book that I describe there.

00:46:47Yes, very interesting, I may jump into one aspect, because basically, if you now

00:46:53look again at society, where we are currently discussing,

00:46:59it is rather the topic that I say, there is a lot of courage and home office is the right thing.

00:47:04What you are describing is that we need to rethink,

00:47:09Is my colleague actually still human or not human.

00:47:13And I have to accept that occurrence and I have to go in that direction.

00:47:15It's also like this, when I feel into Germany,

00:47:20not everything is bad.

00:47:21I also don't want to turn it negative and make it a farewell to Germany.

00:47:23I believe we are at many things,

00:47:25I briefly also talked about it at the World Design Witty Day,

00:47:28in a kind of pendulum discussion as well,

00:47:30and I also talk about the AI act and other topics,

00:47:33It's not bad what we are doing there.

00:47:35We might actually have a fund in our hands,

00:47:38that we approach certain things with a certain,

00:47:41let's say regulated mechanics at one point or another in that moment.

00:47:45But we should be careful that we maintain this courage in other areas,

00:47:48and not constantly

00:47:52rethink whether things were better back then,

00:47:55but rather how we can learn from the topics we have done,

00:47:58from the mistakes or from the things we did well and say,

00:48:01Yes, this new world will be completely different.

00:48:03This is a hybrid world, which does not consist of us having a hybrid meeting,

00:48:07but it's hybrid because we will be working with machines in a manner and way,

00:48:13that we probably cannot fully imagine.

00:48:14And that is a question for me to raise in a counter manner.

00:48:19When is it going to be that we will talk about an Agentic Only Enterprise?

00:48:26For how many years are we talking about?

00:48:27I would say, if I dared to guess, it would be Elon Musk.

00:48:31Elon Musk says, I won’t work in a few years anymore.

00:48:34We also don't need to earn money anymore.

00:48:36We are on a different professional landscape level and because the tasks

00:48:39will be done by robots and by physical and virtual robots.

00:48:45But if you think it through to the end, it’s not so obvious.

00:48:49Because for 80 to 90 percent of the activities, if we’re honest

00:48:54and look at the administration in large companies.

00:49:00I’m quite sure, it will be Darkroom.

00:49:04But that doesn’t mean there won’t be people involved anymore.

00:49:07The role will of course change.

00:49:09And the role will already be that expertise will be important.

00:49:14I know what I’m doing and I can further develop it,

00:49:18which I haven’t done for all these years.

00:49:19All these years I have feared systems, just typing

00:49:22and processes following. That was my activity. 99% were standard processes,

00:49:29that’s what they wanted from me in SAP and data entry. And where the interfaces were,

00:49:35between systems, I maintained them by simply throwing in the data.

00:49:39That is gone. And now I have the time and the opportunity,

00:49:44to prove what I’ve always wanted in my job,

00:49:46when one started 30, 40 years ago. Relatively, without ideas, trying out, new concepts.

00:49:52Yes, let me directly ask a question, because that's the tension. What is this

00:50:00being an expert? Because someone mentioned an expert in faith earlier? Is that still

00:50:06the Data Scientist or is it rather, what is this core competence of ours, that we need to express,

00:50:12as humans in the future when we deal with

00:50:15agentic systems. Is it then actually these soft skills that

00:50:20one used to call creativity, is that

00:50:23more and is it not so, because you just said, I have this

00:50:26expert insight and I believe, so my personal opinion was, as said, happy to hear

00:50:30yours, is rather this general humanity and dealing with systems,

00:50:35that do not react deterministically, but actually really

00:50:39fluid, creative, that may have done something yesterday that they do today

00:50:45completely differently and reacting to that. I believe that's my view and

00:50:50I would love to hear your opinion. What makes us humans, I believe, experts even in this

00:50:54whole system. So this high adaptability, more than knowing how, I don't know,

00:51:01how to best solve a topic in the job script or something like that, or

00:51:05do I want to optimize the code again. Because I believe that's not going to be

00:51:07in demand anymore, is it? So what is the core competence of the future, honestly?

00:51:11So I’ll put it this way, I recently spoke with someone who told me,

00:51:16knowing how to code in a programming language will become irrelevant because the machine writes the

00:51:23code and if humans should ever take a look at it, that's sufficient if it's, I don’t know,

00:51:28something, but you don’t have to, I advocate Java, I advocate Objective C,

00:51:33I advocate it. I advocate something. That was one thing. The second thing that I found really exciting,

00:51:38the topic was that machines actually don't care at all. Whether they look at source code

00:51:43or binary artifacts, meaning the result type. The idea being, here are two result types. The

00:51:48machines will read that as needed and say, here, I have a delta and here

00:51:52is a small report. What happened between the two releases? In a good scenario,

00:51:56it was said, what happened before the customer and what happened with the bug fix? In a bad scenario,

00:52:01said, the following security vulnerabilities have emerged recently, have fun using

00:52:03cars. From that perspective, I do believe that job profiles will all

00:52:08change. Yes, Gardner also recently wrote again about how quickly job profiles will

00:52:11change around 20,

00:52:1528, which made me think, okay, some strange numbers floating around.

00:52:19From that perspective, I do believe that things will change,

00:52:22but, since you’re asking what stays with us afterward, with people? I would

00:52:27definitely say that what stays with us is what you created as creativity, this being both in the

00:52:36some expert roles, but on the other hand also bringing the creativity that the

00:52:42machine, parenthesis not, parenthesis can’t. So this out-of-the-box thing. So this,

00:52:49I was just appreciating again recently from Nanoban, from Gemini, yes, Nanobanana does,

00:52:53for example, also sketches from houses, then the 3D visualizations, where I then thought,

00:52:58okay, construction drawings and similar things. At this point, then making this adaptation, a

00:53:04technology comes, a possibility arises, a procedure comes, an agent comes, something comes

00:53:08around the corner to say, where does this have an influence, where might we need to

00:53:14provide this to the teams we talked about earlier, that come together dynamically and

00:53:18solve a problem. How can we make that available to them? I believe,

00:53:21The human will still play a significant role, but in this topic, we'll code it down,

00:53:27let's connect a database, we won't know here, ABC, so what you would still like to call the

00:53:32routine activity or something like that. I believe we will notice it faster than

00:53:37shorter that the machine is, let's say, standing by at half capacity

00:53:42and supporting you. Let's say, because we just learned about formats,

00:53:47that we will record another episode in a year, because what we talked about today,

00:53:52probably...

00:53:53When is the book coming out?

00:53:54When is the book coming out?

00:53:55Is there already some kind of trend?

00:53:57So...

00:53:58In January, the little paprika?

00:53:59Okay, January 2026.

00:54:00Okay, very good.

00:54:01Looking forward to it, and then we'll talk again in the fall of 2026 and have to

00:54:05tell many things anew, that we are slowly wrapping around the famous loop

00:54:09again.

00:54:10I believe we've discussed a few good topics today again.

00:54:14I would like to ask just one last question, because I considered myself earlier as

00:54:20a half-IT lab autotab and how much about this topic skills, competencies, now

00:54:27as an organization or as a hiring person and you just mentioned the

00:54:32people around us, our children, people outside who are just at the beginning of the job market

00:54:37and so on.

00:54:38What would be the advice?

00:54:39Is it now study IT or study social sciences?

00:54:43So reflexively, one always says in this situation, what the human does best, everything that is creative,

00:54:49you go into the social topics or in, I think probably exactly the opposite.

00:54:55I would really advise to follow the Meinrad, don't follow the trend, do exactly the opposite.

00:55:01If everyone tells you that the machine will do it anyway, yes, the machines will never be alone.

00:55:05So, I would tell my son, one is distributed there right now,

00:55:10is to dive into the AI topics and help shape the AI themes.

00:55:16So, not only use tech equity, but think about what am I building there?

00:55:20Develop creativity with the tools.

00:55:24And what kind of product can I build, what kind of side project,

00:55:28I know concepts, I build you challenges and set up different rates,

00:55:33besides having your own consulting tin, that's great.

00:55:37And my son, who is 13, I would claim that he is on the right track,

00:55:40studying mathematics, because with that he can do everything.

00:55:44And the logic you learn can then be used everywhere,

00:55:47because no one knows what it will look like in five years.

00:55:50And nothing against social sciences, they are also needed.

00:55:54And I love art and I love literature,

00:55:57that will continue to remain.

00:55:58But we talk about them when I am in the Edis.

00:56:00We talk about people who are marching into the normal careers, how they react

00:56:06should.

00:56:07Yes, then we should do it again in the year Timestank and just compare

00:56:13in a year, what has changed and can we answer the questions better then?

00:56:18Yes, we will definitely laugh and we will answer them better

00:56:22can.

00:56:23Definitely as well.

00:56:24I think that can stay as it is.

00:56:25And I also find this point and now we really make the last

00:56:29sentence for this episode. What you're saying right now, that with the topic your son should, feel free to study mathematics

00:56:33When that chatsy-petit moment happened, I actually started with the basics

00:56:39to read through the physics. Because I believe that it always refers back to,

00:56:45the fundamental laws that we have, whether they come from biology, chemistry,

00:56:49mathematics, or physics. Yes, I believe it will come down to that.

00:56:53eventually reduce again, because those are simply the factors that are just

00:56:57there and from Mark AI then discover things that we still do not understand, but

00:57:00I think that's probably not such a bad tip you just posted out there.

00:57:05So reflect on the fundamental laws of the universe, try to understand them,

00:57:12try to understand people, because I think that's always important. I also believe

00:57:15that this whole aspect is simply important, and yes, I'm looking forward after this episode

00:57:20not one bit less to the future with AI, because of how you hinted at it

00:57:25that will remain exciting, that it will be challenging, that we, I believe,

00:57:29all have to stay engaged when we think about how the enterprise of the future

00:57:33looks.

00:57:34I wish you lots of success with the proofreading of your book.

00:57:38We would be happy if you let us know again at the end, then we will also add the

00:57:42link retrospectively to the show notes, so that people who

00:57:45might listen to the episode in the year 2026 can directly access the book

00:57:49and read what was said today, what we

00:57:52missed.

00:57:53The invitation stands. We will see each other again in the fall of 2026.

00:57:57Thank you very much. Great.

00:57:58Yes, I will send myself the invitation from the site. Thanks.

00:58:01I am also looking forward to that when I get to read it.

00:58:04And from that side, I would say to the audience,

00:58:07if you liked it, leave a like, leave a comment.

00:58:11If you didn't like it, feel free to write us an email

00:58:14and spread the word, so that even more people might hear

00:58:18what you liked or didn’t like.

00:58:19From that side, Martin, thank you for being here.

00:58:22We invite you again. We look forward to the book and at this point. Bye, see you soon. Bye.

00:58:52Cardi for thinking, smiling, and above all for discussing.