AIAW Podcast
AIAW Podcast
E138 - High-performing enterprises in the AI era - Rainer Deutschmann
In Episode 138 of the AIAW Podcast, Dr. Rainer Deutschmann, Co-Founder and CEO of L&R Holding GmbH, delves into the concept of high-performing enterprises in the AI era. Drawing from his extensive experience as a Group CxO and former COO at Telia, he explores how organizations can be viewed as living organisms, requiring adaptability and resilience to thrive in today’s volatile environment. The conversation highlights the importance of building and evolving a robust Data and AI foundation and infusing value into all parts of an organization. Dr. Deutschmann also shares his thoughts on the future of AI-driven enterprises, touching on topics such as artificial general intelligence (AGI) and its transformative potential. The episode weaves in discussions on leadership in the tech world, personal health insights from MRI brain scans, and the role of large language models in leveraging private data as a competitive asset. The dialogue underscores the need for leaders to balance detailed product focus with strategic foresight, embrace curiosity, and foster a commitment to learning in a world increasingly shaped by AI innovation.
Follow us on youtube: https://www.youtube.com/@aiawpodcast
into the morphology and, you know, derives quite some interesting things out of that brain scan. It's an MRI, so you actually do quite some scientific exercise and in my case I can say that I learned I need to do more yoga. I'm younger my brain is actually five years younger than I am, which is a good news.
Anders Arpteg:So I'm very happy about that.
Rainer Deutschmann:So I'm younger than in my brain, that I'm looking, maybe, but I can improve on yoga. So that was the finding for me. So but.
Henrik Göthberg:But so Douglas, was he the founder of Voi? Yeah, co-founder.
Rainer Deutschmann:He's one of the one of the key people Still co-founder, yeah.
Henrik Göthberg:And now he has a new venture. Now he's doing a new venture and now he okay. So there are two cool things here, like first of all you have an event on the brain scan and then in the bigger picture here you're trying to yeah, we talked about it a little bit Like you're trying to sort out a community here around.
Rainer Deutschmann:I guess applied AI type projects yeah, the broader topic, at least that we are interested in, my wife and I, is indeed AI and data, and then looking at the verticals, the industries and how to basically get value out of the technology right. I mean, that's what we are after, and it's a very dynamic field. Needless to say, a lot of big guys, but also quite some smaller guys using the big guys platforms, and that's where we will see a significant amount of innovation and value, I guess, going forward. So that's exactly what we are after.
Henrik Göthberg:And just without maybe giving away your idea here. But you're thinking about this a little bit like a community. That's what I get out of it.
Rainer Deutschmann:No, it's no secret at all. In contrary, I mean, we are, of course, advertising. What we have seen is there's quite a number of events, typically where people come and mingle. The issue that we have seen is sometimes they remain a bit superficial. Not saying that it's a bad thing, but what we want to complement is basically more an intimate setting with fewer people that get to know each other pretty well. So it's everyone's actually introduction. What is the investment thesis? What are the stakes that we have taken? What's the kind of the ticket size? It's really a bit more deeper and then we engage in a conversation is not a presentation, but it's actually a very intense discussion and yesterday we gave douglas a hard time because there is, of course, questions on the business model and how many people can afford an mri scan, and those kind of questions. How big is the market?
Henrik Göthberg:so, so it's going deeper, yeah that's the idea and is that everyone is so curious about this, either on the ai startup side or on this sort of family investment side. How do they sort of get in contact or become part of that I?
Rainer Deutschmann:think it's really going through the contacts that everyone has and we are generating typically a curated list of participants that are specifically interested and relevant for that particular topic. So that's why I mean it's like 15 people max about this is what I meant.
Henrik Göthberg:It's not sort of super open Open is in the sense we know it's there. But if you're interested in participating in stuff like that Contact me.
Rainer Deutschmann:This is probably the first time we have any kind of discussion or advertising about it. There's nothing known about this.
Henrik Göthberg:No, but this is a contact-driven approach.
Anders Arpteg:Yeah, exactly Is no, but this is a contact-driven approach, yeah exactly Is this you and your wife that drives some kind of investment firm.
Rainer Deutschmann:It's a tech arena people as well. So I mean we have an investment firm indeed, my wife and I since quite a number of years, but we do this together with the tech arena, and Omid and Michaela are the two people there, so it has some weight behind what we're doing.
Henrik Göthberg:Yeah, yeah, but you started a company with your wife. Yeah, when was that?
Rainer Deutschmann:That started humbly a number of years ago. Even the start date doesn't matter, because I mean, we started, as probably everyone, to look into interesting more friends and family type of innovations and startups. And then we quickly discovered especially my wife is doing this full time that this is a full time job. You cannot expect to invest and generate meaningful portfolio on value if you do it on the side. It just doesn't work. It is what probably quite a few people think I have a friend and I invest a few thousand dollars, but that's not really what in the end will work on average. I mean, maybe there are some lucky strikes, but you've got to be building a bit of a much wider lead, I mean lead funnel. And then there's about you know what are the picks, who are the co-investors, who is the lead investor and so on. And then, of course, how long do you stay If you go early stage? Are you going to be living for a while or are you going to be?
Rainer Deutschmann:selling it off in the secondary. So it's a bit of a and of course it's quite a bit of an ecosystem which is not like accessible just if you come in from nowhere. So it takes a bit of a couple of years.
Anders Arpteg:I'm also in the investment arena a bit and it would be fun to hear what's your. I'm also in the investment arena a bit and it would be fun to hear do you have a favorite kind of way to gauge and evaluate different companies?
Rainer Deutschmann:Yeah, that's a huge topic and this is exactly also what we discussed, for example, yesterday. What's the investment thesis and so on. Typically we feel as a smaller investor that we will not be able to do like a 360 due diligence as normally a vc would be doing. So that's why one of the things that we are doing is we are quite leaning in with larger top vcs typically. So that's one of the things.
Rainer Deutschmann:We would, of course, look at the typical, I mean basics, like the product market fit. I mean, is there like a demand, that is, the, is there a team that is capable of, you know, serving that market? And I think one of the most important things, even that was unanimously yesterday quite agreed with the other investors it's really about the founder or the founder team and, interestingly, if you look at the characteristics of founders, the ones that have gone through very tough times, even privately, typically are the more successful founders. It's very interesting. So, if you have a challenged childhood, if you have gone through hard times, people that have really seen the dark side, I would say, have a bit of a resilience and they get up and they will more likely than others not guarantee, but they will more likely than others be able to drive it through.
Rainer Deutschmann:It's actually a little bit. There's some truth there.
Henrik Göthberg:But you know what? Then I think we can make a deal now that we need to have your wife on the pod here, maybe with Douglas or someone else, and we can dig deeper into this community topic and how that has been progressing. Maybe that would be good. And then with that, I think it's time to introduce you properly, dr Rainer Deutschman. You need to say when you have that I have worked with. That was a joke, but it's not a joke.
Rainer Deutschmann:It's actually interesting if you talk about doctor, because if you have worked in a couple of parts of the world, there is actually regions in the world where it's been taken more serious, whereas in other parts of the world, like probably the Nordics and Sweden, it's rather strange if you even say the titles yeah and I worked with, Don't worry about it?
Henrik Göthberg:I don't worry about it. I worked with Watt't worry about it? I don't worry about it. I worked with.
Anders Arpteg:Vattenfall in.
Henrik Göthberg:Germany. Yeah, the doctor was quite important. Yeah, so, dr Reiner Teutschmein, I want to do it properly. You are so much welcome here today, thank you for having me.
Rainer Deutschmann:It's really amazing.
Henrik Göthberg:Yeah, and I mean like I'm just thinking you had such an interesting career and we don't want to spend 15 minutes, 20 minutes on your career. But there is an interesting story here. You have a research background in physics, developing early quantum computing stuff. You had a background in McKinsey strategy and then you moved into a long, distinguished career in Deutsche Telekom, all the way up to senior vice president level in product and then from here reliance, jo, exata and now what maybe it's most known to the swedish community you are the former group uh coo for telia, and then you just left that position, I think in june this year correct, correct, so okay.
Henrik Göthberg:so it's very exciting to have you here, and maybe you know who is Rainer Deutschmann today, or you know what would you like to sort of highlight of these?
Rainer Deutschmann:You know we can go on for hours on this, but you know a little bit so we can understand who you are in a couple of minutes. If I just summarize, or just try to craft it into one word, it's curiosity. I think it's really what drives me, since and I think it's so important for all of us when we are kids, we are just so open-eyed going through everything in the world, and then, over time, we seem to be a little less open-eyed, and I think that's the one thing that I really want to maintain. I want to just be absolutely curious about new things, about what's happening and obviously, what you guys are doing. I mean, it's like looking out and finding the interesting nuggets that are happening in the world and then how to really leverage them.
Rainer Deutschmann:So the curiosity is the one fundamental trait that I would describe myself and in terms of more business, if I craft this into a business purpose, it's really how to use, or how to enable people to use, technology for value creation. That's really what I'm excited about. So not just technology for the sake of cool stuff. It's really how to amalgamate teams, individuals, teams, larger organizations with the right technology to create value which can be revenue, which can be profit, which can be cool. Products and services, customer experiences that's really what I'm excited about, more like from a business perspective.
Henrik Göthberg:And I think we can use. There's so many experiences here throughout here that I think when we are talking about the main theme, we should try to draw out real examples from the different uh parts throughout the career. But if, if you would now summarize, I mean like if, if I look at you today and we are inviting you as a guest, I mean like you, you are, in my opinion, a big boss, right so? Group ceo of telia. When you came out of that job, so how would you describe your leadership style?
Rainer Deutschmann:So obviously I would just love to have the curiosity in everyone. So I really try to inspire, I try to get people open their eyes and discover, of course, the problems that are there not closing eyes for problems, but then also really look for the right solution. So that's really what I try to just scale beyond myself. Actually leadership and you should ask some of my team members if they can confirm I think they hopefully do is that I, um, I can, uh, I can, I can, I I always am with the team when there is a need. So so that's that's kind of what I'm, what I'm really excited myself. So I would never let anyone down, especially if there's a hard problem and, um, whether it's designing a strategy, whether it's designing a specific solution for a problem, a customer problem, a cost problem, whatever it may be, a transformation issue, a growth problem, how to tackle it, but then, of course, enabling my team always to then run and scale by themselves.
Anders Arpteg:So that's a little bit what I'm really after Like a role model of like a top C-suite person in some company throughout the world that you look up to and say this is a leadership style I really admire.
Rainer Deutschmann:I have seen, and that's actually even almost like a choice for me at least where and with whom to work is really the people in the organization and the leadership team and the boss, so to speak, rather than the specific organization itself. I can say. Give you one example you mentioned Deutsche Telekom. I was there nine years, which is a long time. Great company, amazing transformation and still one of the best I would say probably even the best telco in the world, when you look at the hard factors in terms of value creation, but also the soft factors in terms of the cool thing DT is doing.
Rainer Deutschmann:But I was then approached by a person called Mukesh Ambani, who is maybe less known to everyone, but he is the chairman of Reliance Industries, which even may not be known to many people.
Rainer Deutschmann:It's a large conglomerate in India and he was and there's a long history to that but he was basically after creating a new telecom operator for India to boost India into the first world of digital infrastructure, so coming from a slow 3G into a new 4G and now 5G network for 1.3 billion people from scratch. So that was the vision Completely crazy, especially crazy given the fact that at the time which was like 2012, 13, 14, 11 other players were already in India, so why would you then build the 12th one from scratch? Right, I went there. He called, and then I just had this amazing mind-blowing session where he put forward the vision for India and I basically got fired up completely. He cancelled the German presence, took my family and then we went to India and then I was part of the leadership team to actually, you know, design and launch that business, which now happens to be the largest data telco in the world.
Henrik Göthberg:And what leadership was he representing?
Rainer Deutschmann:That's the point I wanted to get to is. So he had a unique combination, which is what I got inspired so much of having a really large vision. But then he paired it with a passion for detail and a hunger for trying to understand everything in that industry, which wasn't necessarily the industry that he was coming from, everything in that industry which wasn't necessarily the industry that he was coming from, because while he had Telco before and that got handed to his brother, that's why he built a new one, but he is more like a person that really generates the cash from oil and gas industry and from retail industry and other industries. So he basically, as a chairman, went into really the depth that you cannot imagine, where everyone of an engineer would be in awe to say, okay, he really understands, and if he wouldn't understand, he would go after and not rest until he would understand.
Rainer Deutschmann:How is 4G working? What are the critical things? How to build a network, what are the vendors, how to fit the network and the handset. How to build a network, what are the vendors, how to fit the net, the network and the handset. I mean gazillions of questions, how to specifically build something at that scale which was never done before in the world. Nobody ever had built a network for 1.3 billion people from scratch. It didn't exist. So having a person with a vision, at the same time, with that passion for detail and hunger to understand everything, that is really inspiring to me.
Henrik Göthberg:But there are some examples of people I mean like we can talk about.
Anders Arpteg:Elon Musk, or we can talk about Sundar Pichai, Satya, Tim Cook, etc.
Henrik Göthberg:And then I take a step back and think about the leaders that we meet in enterprise and how many people that really are that wants to be strategic but also have the stamina sort of to deep dive and go underwater deep. I think that is an interesting observation you're highlighting and we can reflect on who is doing that.
Rainer Deutschmann:Yeah, I think not enough people. I think there's always this management discussion, honestly, and how much should you really take on yourself, how much should you delegate? And that's a very tricky question, but it's just. You know, those examples that you mentioned and the one that I personally know, which is Mukesh, is just showing that, yeah, great leaders, I think, have a passion for the actual product or service of technology they're dealing with. I don't believe in leaders that would just come in high level and then try to manage.
Henrik Göthberg:That's not my leadership style and that's not what I really believe in and if I steal your argument here, it's like oh, you're talking about a micromanaging leader and I argue not. I argue micromanaging isn't something. But this is about understanding, and in depth understanding. How do you make money, or where does the money flow? What happens? Why is this a problem? It doesn't necessarily mean you micromanage, my opinion. So I think it's a very this is a misconception that people see. When you're interested in details, you're micromanaging.
Rainer Deutschmann:I totally agree. So I see it more as a partnership, as a sparrings partner, to be able to problem solve with the team, not letting the team alone at any point in time and being part of the solution, but of course then you know letting go if the things work. I mean, of course, a leader like again Mukashi would have quite a number of people that would run and he wouldn't have to focus on, but if there's a problem you would know that. You know we are in there together. So that's, I think, a good leadership style.
Henrik Göthberg:And you can also flip it in the sense that when the problem is at its deepest, that's when we wanted the manager or the leader not the manager, the leader as the closest. Yes, so it's another way of looking at it. When we have a problem, I feel accountable for this problem, together with the team that has it. Yes, exactly, wow, but anything else, you know, because I want to leave the introduction. But I know me and Anders has always found Anders being with a research background and it's like, oh, we need more techie guests and if there's any small ounce of techiness or nerdiness or so like, we need to jump on it. So maybe one or two minutes on your research background quantum computing, there's some cool stuff in there, yeah, Computing neural computation.
Rainer Deutschmann:I think I was lucky, um, based on maybe the curiosity. I was lucky that I had the kind of opportunity to look into a certain number of fields which later became quite, I think, more prominent. And you mentioned one of the things I was. I was um in a as a physicist there's two types of physicists, by the way right the theoretical physicist and the experimental physicist, and I don't think anyone is better than the other. But I was one of the experimental ones, actually doing the lab work and suffering through all sorts of problems that are beyond the pure theory, problems that are beyond the pure theory. And so one of the problems we looked at is what happens if you are shrinking structures and putting electrons through them at cold temperatures, at some certain magnetic fields, and what are the effects? And the quantum effects, so it's called quantum transport, and so those kind of things kept me busy, and I've done of course now it's at much larger scale the question of quantum computing. At the time it wasn't really at all even thinkable that you would start to, I mean, potentially overcome the noise problem, which still, I think is one of the key problems too, which is not totally clear if it's ever overcome. But I mean we are getting probably closer to understanding.
Rainer Deutschmann:The other, very different field I was looking into, also experimentally, was the neuron silicon coupling. So now, of course, what neuralink is doing? Um, which is absolutely mind-blowing. I think one of the it was quite prominently discussed when yelon launched it, but now less I see news. But it's absolutely amazing. And what's the? How can we, how can we increase the bandwidth between us and the computers? We are like so low bandwidth by talking, like what we are doing now, by looking at screens and by typing? It's exactly, it's like ultra low bandwidth for any standard. So how to increase that bandwidth is a key question. So I was also myself plugging neurons on a silicon dioxide field effect transistor, trying to measure the action potential or even initiate an action potential by having the field effect transistor being active. So this kind of stuff I've done. And the last one to mention maybe is then also neural parallel computation, which is called neuromorphic engineering at caltech, which you're smiling we can throw out the script now?
Rainer Deutschmann:we should maybe stop there because? Because we can have an org discussion on pneumorphic versus quantum and where to put investment yeah, and it's interesting because some of the people that invest, that invented the transistor at Caltech, carver Mead and some of even his doctorate students, later started this operation of the transistor in the nonlinear region Right. So we normally use the digital one on off, but the neuromorphic engineering that we have done is looked at the whole exponential curve of the transistor and you use that for-.
Henrik Göthberg:It's time to leave the von Neumann architecture Potentially. Yes, I'm an out of this list.
Anders Arpteg:I have a big passion both for quantum computing and especially for neuromorphic computing. You're not going to get out of this seat.
Henrik Göthberg:We can turn off the camera, but we're going to fool you up with beer and we're going to talk all night if you want. Anyway, but interesting canvas, because now we're getting to the theme of the whole, you know, of today and we had a conversation before and we were thinking about different themes and, you know, playing on AI and technology and also the fact what does it mean for the large enterprise, and we were flipping a couple of different topics and in the end, I picked one of four ideas you had which was the one that I liked the most, which we called high-performing enterprises in the age of AI, and I really liked it because it highlighted not only AI or AI-driven, but it was saying something about high-performing and AI. So maybe we start there how would you pick apart the theme? How do you define high-performing and how do you do that in the context of AI? So how do you pick apart the theme?
Rainer Deutschmann:Yeah, I would maybe pick apart the two elements in terms of. One is to my earlier point. Now we have obviously AI at the next level. We had AI in machine learning since a long time, obviously, which is it's got a long, long history and sometimes it's a little bit forgotten. But now the Nobel Prize has at least surfaced a bit of the history forgotten. But now the Nobel Prize has at least surfaced a bit of the history.
Rainer Deutschmann:But I think the current state is so that we need to again make sure that we enable people with the right technology to deliver value in the organizations, and I think AI is one of those where that's a lot of work to do that. So that's the one part and I am very excited about seeing that take fruition in the various elements of applying AI in the companies that I was working with and the other one is the question on high performance, as you were asking for the definition. I think for me, high performance in the end has probably two aspects. One is it's all about value in the end. Let's face it, if we are running companies, we need to deliver value and typically we look at shareholder value and then, of course, underneath there's the value for the customers and the values for the employees and typically, the societies that we operate in. We've been quite clear, as Talia also, on these dimensions.
Rainer Deutschmann:The other one in terms of high performing for me is also a non-static dimension, which is the resilience part, because I may be high performing today, but that's not good enough. I need to have capabilities that allow me to overcome unforeseen foreseen, of course, but also unforeseen issues that come. Maybe that's the inflation going through the roof, which we had a few, I mean since a few years. Maybe it's a new technology coming in. Maybe it's a new competition coming in. Maybe there's a new regulation coming in which is specifically important for telecommunication. So all of these things are having the I mean there's this risk to derail an organization that may be high performing today. So I think for me, high performing is value, but it's also resilience and being able to learn and adapt faster than the competition.
Henrik Göthberg:In the end, this learn and adapt faster. Leading up to this podcast, I watched some of your own presentations you've made that you can find on YouTube and I think this is from 2018 and 19 is in Sri Lanka, where you highlighted to the power of X, where you highlighted the fundamentals everything is exponential and the reality that we, as people, are quite linear in understanding yes, yes, all brain is not really made for.
Rainer Deutschmann:And then?
Henrik Göthberg:you know you started in that presentation on. You know exemplifying with morse law, and we can add netcal flow, we can have a curfew law of accelerating returns. Everything points to an extreme exponential topic, which then means what is high performance in an exponential setting? And how do you flip that?
Rainer Deutschmann:That's even more important, I think I mean that's a very deep question, especially if we look at the nonlinear effects of a compounding technology and what that means as an example is, obviously, if I have systems that we have now as a, let's say, large language model foundation and I'm starting to use the technology to optimize itself, I mean that's basically a hyper exponential behavior which I haven't seen that much yet, but I'm sure we will see more of that. And the question is then if I am having a number of market players, isn't it so that in the end, you know, one of them will have a little advantage and that advantage exponentially compounds and therefore you will have an absolutely, you know, massive advantage, uh, which is which is resulting from that exponential compounding effect? So so that's uh. We haven't seen that.
Rainer Deutschmann:I mean, there are the big guys out there, but at least we don't have just one. We at least have uh on the on the llm side, we have open ai, we have anthropic uh, we have germany a little bit fighting coming back maybe, um and we have OpenAI, we have Anthropic, we have Germany a little bit fighting coming back, maybe, and we have, of course, the Facebook with Lama. So those are the four, it's not only one, but I mean, is there going to be a world where we'll hopefully not just have one is a question.
Anders Arpteg:I think that kind of gap, if you call it that, between the top giants, the four, or I think you should include xai as well there.
Rainer Deutschmann:But yeah, they might get this they claim to be the biggest one in the end of this year.
Anders Arpteg:We'll see, that's elon scale, by the way. Um, but do you think that kind of difference where we see the top giants really exploding and being the most valuable companies in the world today and the rest of the companies not so much and not really perhaps expanding in value in terms of monetary value in this case as much as they are, do you think that kind of divide will increase or will it actually start to decrease?
Rainer Deutschmann:that's why I honestly don't know. Um, I think there is currently and we didn't at all pre-discuss this to to to make a topic, but I think there's a huge discussion, as a physicist, quite interesting what's the scaling law? How will the performance of those large language models scale with things like the computation power, with things like the amount of data that you actually use? And as we, I think, know now, we have pretty much exhausted at least the publicly available data, so you're not going to be getting more training data from just web scraping and so on. I think the big frontier now is the private data, and there's obviously orders of magnitude more which is hidden in some company's database and in the behavior of customers that is being monitored, and synthetic data. I mean, those are the two which haven't yet been fully, I think, leveraged for what you are asking in terms of going to the next level.
Anders Arpteg:If you take like AlphaGo or something, I mean they purely operated on synthetic data. Do you think LLMs in the future will do that as well and they can simply generate their own?
Rainer Deutschmann:I would say AlphaGo Zero at least did not just use synthetic data, but you could. I mean it's basically starting to play against itself. I think that's going to be the next level where, instead of you know, us humans doing human, I mean human feedback, I mean the reinforcement, learning through human feedback that is, of course, still probably required for some, let's say, alignment problems, but I think the human will less and less be there because it's just so much stuff to do we cannot scale. It will be, I think, what we see now emerging with the reasoning engines, where the model itself starts to ask the question to itself himself herself, whatever we call it, and reasons through a certain problem and then also checks it back before the answer is given back to the user. So this we'll see more. So it's a system to system rather than a system to human evolution, it's getting more.
Anders Arpteg:I mean AlphaZero, as you say, is purely synthetic, but AlphaGo at least was a combination of the two. Yeah, I agree, and that's similar to LLM01 model of today, right and used to throw some fuel in the fire here.
Henrik Göthberg:Then we have all the articles coming out on the deterioration of models, when they argument that if we are not having any real human data, you know where do we end up here. But bringing it back to the theme, so we are talking about a world order, so to speak, high-performing in a world order that is quite VUCA, volatile, uncertain, complex, ambiguous, and we are dealing most likely with scale effects or exponential effects. I mean, like everything points in that direction. So high performing. Do we have an organizational template that is suitable for that kind of exponential approach?
Rainer Deutschmann:I think, linking to what I said before, I don't think we should look at organizations as static organisms. I think we should look at organizations as living organisms that have a past and a future. And then the question is what are my features that I can use to evolve better and faster? In the context of commercial organizations it's obviously then against competition and to make market share and so on. So what are my capabilities that I need? I think I would look at a more optimal organization around those features and those capabilities, and some of them relate to what we said earlier the kind of the learning mindset in the organization. Do I have people that really are open and curious? I mean that's so important. Do I give room for experimentation and failure, right, is another big thing. Do I learn from those failures? So, in a way, do I codify and put myself on top and then leave that behind? Or do I make failures over and over again, which does happen in?
Henrik Göthberg:some organizations Feedback loops, single loop learning, double loop learning, all this.
Rainer Deutschmann:All these things. So I think these are elements, and it's not a holistic description, but these are elements which I do think are important for organizations that we need to foster and build.
Henrik Göthberg:But now you guided us in a really nice segue from high-performing enterprises you elaborating, putting a canvas up on the theme into a topic that we were planning to discuss, and and there was your idea of thinking about an enterprise as an organism. And you already did the first steps here. And the segue is very important because it it contextualizes organization as an organism, not in a vacuum, but why it needs to go in this direction and it has to go in this direction and it has to do with all this VUCA stuff, this exponential stuff. You can't know everything, but if we do that, we can move into another question, because when we discuss this, if we have an organism in front of us, then we can think about this how to fix it, how to improve it or how to know what to improve. We can start metaphorically understanding it. Well, any organism needs how do we check its health? How do we check it's good and all that Correct? So how to think now about an enterprise as an organism? And you continue spinning on this yeah, so.
Rainer Deutschmann:So it is a, it's a thought. It's like any any uh picture. It will have its its flaws and probably shouldn't drive that analogy too far, but I think there are some truths in comparing an organization with an organism. There are many parts that need to work together. Right, there are many parts that are connected.
Anders Arpteg:If I'm trying to understand that, I mean who doesn't think an organization is dynamic. I mean any organization at all right, I agree. So is there something we should? Or just help me understand what you mean.
Rainer Deutschmann:Yeah, I hope that nobody will look at an organization as something static, but I'm also not 100% sure whether this is true and what does it really mean? What are the features like self-repair? Coming to this organism analogy, do we have antibody against issues that are coming from outside?
Anders Arpteg:Do we have, like, a biological organism in the sense that it actually has some kind of resilience, that is what I'm thinking about Exactly.
Rainer Deutschmann:Even if you take a human body, I mean you have a brain, you have arms, you have legs, you have organs, you have a spine, you have neurons, so all of these things. Actually, if you think it through half, I mean you can make an analogy to an organization, a commercial organization, and it's really about recognizing that there's quite a number of different elements required for the overall performance and it's very living and many things are transported through the organization. I mean you can talk about the neural signals, of course, when the boss or the executive management is, you know.
Rainer Deutschmann:Your data decision flows, and then there's data, and then there's blood, which is like transporting energy. That could be the money, so there's, I mean you can spin that forward if you like. But the point I think, as you say, the money so there's, I mean you can, you can spin that forward if you like. But the point I think, as you say, is really there's a dynamic uh element in this and a very um, it's living. I mean, for me, organizations and that's makes them exciting is, is there's life?
Henrik Göthberg:in there and I I can flip. It's how you know, don't we all think this is dynamic already? Yes, we do. But if we think about it, take a step back, the way we look at big organizations and the way we deal with this. We are trying to draw an org schema first and then basically we set that. And if I go back, how has enterprise worked the last 50 years? We do an org schema and then we do a change. So we are not really that dynamic. We are more a little bit like static. We understood something that was slow moving, we optimized one process here and then we do that over time.
Anders Arpteg:But still, if you actually strongman that a bit. I think a lot of organizations at least a lot that I've been involved with are reorganizing themselves very frequently, too frequently, I would argue, even to an extent that it's actually hurting the company quite a lot.
Henrik Göthberg:No, but let's okay sorry now. Now you're going to hear when we start. You know, forget about the guest? Almost no. But the argument here is organization is emergent, it's schema. Second, it's not schema less. It's not that you can't have an understanding for modeling the feedback loops of an organization, but they are emergent. And do we really treat it emergently? You worked at Spotify for many years. You were treating this emergently, I would argue.
Anders Arpteg:It was a lot of reorganization all the time.
Henrik Göthberg:That's not how Vattenfall works, but still I think it was hurtful in many ways.
Anders Arpteg:So I mean you have to do it if it grows a lot.
Rainer Deutschmann:But what do you think? Yeah, I think you're totally right. Reorganizations before or after the reorganization is. Before the reorganization, almost everyone in the organization itself will say, hey, the next one is coming. And why is that?
Rainer Deutschmann:And I think what we need to realize is that we should be looking at capabilities that our people have and capabilities that certain problems whether it's a new product or any other thing that needs to be done requires, and what we need to provide in an organization is the match between the needed and the available capabilities, and no two problems will have the exact same setup, and therefore this issue around trying to organize happens, and I can say back to this Reliance, geo adventure and the learnings there, we have had the same exact issue in the beginning, and then we started to look at how do we set up the organization in a what you now, of course, call a cross-functional way, whereby you just don't look at a line organization as the dominant axis, but you look at actually processes as the dominant axis. So what we have then done is we have started to look at what are the processes that we need to actually conduct, and obviously one of the most important ones is the customer processes acquire a customer, onboard a customer, you know, provision the product, be able to bill and then be able to help if there's an issue, and so on. So those are the steps, typically, that you look at and what are the capabilities required in the organization to provide each of those processes to the customer? That is the way how we looked at it, along a process dimension, and that, I think, is the probably probably better way, because then I can say what are the capabilities and then I recruit the capabilities. Whether it's an engineer, whether it's a UX designer, whether it's an IT person, front and back end, whether it's an ops person, whether it's even a finance or legal person, those capabilities will then be recruited from the respective line organization.
Rainer Deutschmann:But that's not the dominant axis in reality. The dominant axis, in reality, the dominant axis, are processes. So that was the learning to your point. So organization, I think, is only a bit of a host, a grid that we give but that hosts those capabilities. The real important thing is what are the capabilities that I have in the organization?
Anders Arpteg:Shouldn't you have a line organization that is rather aligned with the processes?
Rainer Deutschmann:For some time. Yes, but then what happens if something changes? Right? You would basically have to then change the organization every time a new process comes, and that's the issue with that kind of static. That's why there is a static thinking oftentimes in organizations, and we should dissolve that a little bit and consider, you know, the capabilities as the more important axis.
Henrik Göthberg:But this core topic, of which one is the primary axis, I think, is a very, very profound, deep question. Because you have the functional view of disciplines or competences, you have the process view. I would argue what is the dominant trend now is to move away from a pure process end-to-end view and even into a product-centric view. We are talking about, if you go into books like Team Topologies and understanding value streams, or Agile, safe and all that, this is very much a product-centric axis. And then you have platform teams that actually are products but further down, supporting teams. So this axis of what is the stable view here. I think this is a deep question in relation to succeeding in these topics.
Rainer Deutschmann:Yeah, if you look out a little bit, I think the view that one can have and I certainly have it is that you have the single person company, because in the end I mean, that's not now, but sooner than we think probably we will be accompanied and we'll draw on co-pilots of any form or shape, and the only thing is I'm the pilot and I have a number of co-pilots that will help me. So do I need to have large organizations? Is really a fundamental question in the first place. And so, yes, now we have large organizations is really a fundamental question in the first place. And so, yes, now we have large organizations. But I mean we should look forward. And what happens if we have capabilities at our fingertip? If I have an exponential ability to do things myself, I don't need maybe that many people, especially if I start from scratch a new company.
Anders Arpteg:It's similar to the Sam Altman thoughts about having one man unicorn.
Rainer Deutschmann:Exactly that's what we're talking about here. Exactly that.
Henrik Göthberg:Yeah, yeah. But, and we actually in the last podcast we were extrapolating it to the zero company unicorn, zero man, unicorn.
Anders Arpteg:Then we had fun in the end. Do you have any experience to share about? You know your previous company, telia.
Rainer Deutschmann:Perhaps you know some learnings you had from that in terms of how to organize that in different ways yeah, um, I think the telia is an absolutely amazing company if you look at it, and I was talking about me looking for people when I join, and that's certainly the number one reason for me to come. It was Alison Kirkby who unfortunately left us and now she's leading British Telecom, but she was certainly the number one reason for me to join. And the second reason has been that Telia is just such an outstanding company of high quality and heritage and also, therefore, responsibility for us in the region. Basically, having invented mobile telecommunication together with Ericsson, I mean, it's just you know who can say that it's starting a complete revolution on this planet, on which everything else is now built. Everything builds on digital infrastructure, if you look at it. So nothing works without having a strong, secure, resilient digital infrastructure. So I think that's the amazing part in telcos in general, but specifically Telia, I think. And so having such a backdrop and being able then to help transform the company, like I said, using, enabling people to use technology to create value, has been, has been, just an amazing uh journey.
Rainer Deutschmann:There is a notion which we haven't also pre-discussed, but there's a notion of compounded complexity in organizations, which I think is one of the things we should be mindful about and a whole, maybe other discussion, is how to overcome complexity and how to simplify. I think that's the core thing and that's certainly one thing that we have found when we started the current chapter, which started in 2020 around, you know, simplifying the company and unlocking the value that is in an organization like Telia to be able to serve the customer better, but also to be able to invest more, because 5G isn't cheap at all, and to be able to fund 5G and there will be a next generation at some point in time. I mean, you need to have that value unlocked. You can't live by just cost savingsaving and dying in the complexity. That doesn't work.
Henrik Göthberg:And now, if we now infuse, what's the data and AI perspective of this organism or with this backdrop? So we are thinking about an organism, an organization, and now we're going going to add data and AI into this mix. How do you see that?
Rainer Deutschmann:So I think one analogy and that's why it holds again, we shouldn't stretch it too far, this living organism analogy. But I think one of the things is in an organism everything is connected, while in organizations we often see silos and specifically we see data silos. Now I'm not even talking about AI, but I'm talking about the foundational input to AI, which is all data. So that's the one thing that also I saw in all my organizations that I was fortunate to work in. So to identify and unlock data is one of the core things that will be important Again when we talk about older organizations more than maybe younger, because if you start from scratch, you would probably ideally not have any data locked in silos.
Rainer Deutschmann:So then I think the other part is on process. If I have an organization which has a huge number of manual steps, where people move information or people process certain steps manually by typing on a keyboard and then reading an email and typing another one on the keyboard and so on, fundamentally I cannot automate such a process because it's not digitized right, and so I think the second big thing for us in organizations is to really drive digitization end to end. We need to have digitized processes so that then I can smartify and automate processes. I think that's the key insight that, also most recently then in Telia, was one of the underlying drivers of our transformation. How do we simplify and how do we digitize processes customer process, product development process, technology process, enterprise processes, hr finance, you name it. So to avoid having people doing manual work in unconnected systems, to connect those systems and to be able then to use data for ever smarter end-to-end processing that's, I think, the core driver for performance in an organization.
Anders Arpteg:And you mentioned a big initiative in Telia then in 2020, it sounded like and you were on some kind of journey to simplify, if I understand it correctly, the processes you had there and, if I understand it correctly, you also identified a number of silos and a number of systems that were rather fragmented or not connected in a way that made it really hard to automate and improve. It Is that?
Henrik Göthberg:a fair description, yeah.
Anders Arpteg:And then you went in some kind of journey, I guess kind of to consolidate systems together, or how did you go about trying to simplify it?
Rainer Deutschmann:Yeah, so that's I say in in. These are my words. I mean, this is uh, it's like I try to think about it. I think there is a kind of a three-step approach, specifically in technology-driven organizations, which have scale benefits, and some of the larger companies we talked about have massive scale benefits. So for me, it's about simplification, standardization and scaling. I call it the 3S in my own words.
Rainer Deutschmann:And for what do I need to do that? And that's the other dimension. I need to do this for my product portfolio. I have typically many, too many, products in organizations, especially if they've grown over many, many decades.
Rainer Deutschmann:I need to do it for my processes, as we discussed. I need to simplify my processes or try to digitize and therefore I can automate. I need to do it for my underlying platforms. I have so many unconnected platforms that are driving those processes and products. I need to do it also for my people.
Rainer Deutschmann:So what are the skills that I have and the skills that I need? And I need to do it also for my people. So what are the skills that I have and the skills that I need? And I need to do it for my partners. You have five P's, so these are the five P's exactly. So I need to simplify, standardize and scale across the five P assets. That's kind of the summary that I give in my own words, and I think that there's some good learnings around this how to do this specifically for products, how to do it for processes, how to do it for platforms, what does it mean to bring people along and, of course, also what's the relevance of partners in such a transformation. But this kind of captures a bit the things that we've done Three S's and five P's.
Rainer Deutschmann:If you understand that, then you're on a good path, I guess understand that, then you're in a good path.
Henrik Göthberg:I guess, yeah, but but that actually leads us, as a segue, into a a part of this topic. So, if we, if we, if we take this backdrop, the five piece, three s's, maybe a large transformation in in italia, or even building something from scratch another key conversation we had when you're going on this transformation or building something, another key conversation we had when you're going on this transformation or building something, there is something in here which can be understood as if when you digitize somewhere here, we need a strong fundamental or a data and AI foundation, and this is the word I think you have used many times. Can we go here and listen? You have used many times. Can we go here? And what do we put? How do we define a data and AI foundation, or how?
Henrik Göthberg:are you thinking about this so?
Rainer Deutschmann:two elements we touched upon just now. One is, as we started to talk about the data and unlocking the data out of silos so that we can access it. I think on that one it used to be so that there was a thinking okay, I need to put all into a data lake and so I make it access. I think that will and is changing. So I think now we are able to just access data wherever it is, so I don't need to have that kind of layer of data lake. And especially AWS obviously they are driving very much this technology so I can live much more around the data without having a classical data. I think that's the one part. The other part very important for the foundation, is the end-to-end digitized processes, and I give you one specific example which I find illustrative. It's ServiceNow. I mean some people that illustrative it's ServiceNow I mean some people that listen in here may know ServiceNow. So it's, I think, a transaction workflow engine that allows organizations to put one step after the other and process data in a kind of a user-friendly way. And I think typically when you look at older organizations, there may be many big foundation because we have done it at Atelier, to save that example, and then large language models came and because of the fact that we have had a connection across the company through ServiceNow, connecting all the IT estate and also connecting the customer domain, where somebody calls and has a complaint and that complaint results in a ticket right Same ticketing system we use for the customer domain as well, as if there's a ticket from a server problem or an application problem, same exact system.
Rainer Deutschmann:Now, because this is one system and it's all connected, I can now very smartly deal with those issues. For example, I can correlate is the customer problem related to a certain technology problem? I can do these correlations that happened. Say, for example, an antenna goes down, a 5G antenna goes down, there's certain things I need to do to restart it.
Rainer Deutschmann:I typically have a person doing that, but if I have the digitized process and I'm putting a basically I call it bot, based on a large language model on top, the bot can observe the person doing a certain task and guess what happens next time. You don't need the person anymore. You can take the person out of the loop. So there's a learning system built in increasingly in our organizations if we have end-to-end digitized processes because we observe the human and we can take those standard workflows over in automation, if you know what I mean. So that is the kind of the closed looping that I was talking about, by learning and improving and taking the human into areas where there's uncharted territory, where we still need humans to do, because we haven't yet really either digitized them or we haven't been able to get enough.
Anders Arpteg:Simply map out the systems you do have in some way and be able to collect the data, I guess in a more holistic way.
Rainer Deutschmann:It's map out and connect. Specifically, it's about connecting. I think the point is we need to integrate the parts of the organism, the parts of the organization of the system, so that I can actually deal end to end with them the organism, the parts of the organization of the system, so that I can actually deal end to end with them. So that is, I think, the work, and that is hard work because all of these systems, they don't have necessarily standard interfaces. Some of them may be legacy, and so it's not an easy thing. But once I have done this, my life is so much easier because then I can start to automate and smartify as I described with the bot.
Anders Arpteg:So that's how you break the silos in some way. You then have data across the system that allow you to find the patterns that you, otherwise, would never be able to see Exactly.
Henrik Göthberg:So if you go a little bit techie and break apart the different tech architectural capabilities we're talking about, here, we're talking about some sort of technology.
Henrik Göthberg:Let's now use ServiceNow to exemplify where we can actually build a workflow for the people, and in that sense we can take a manual or an analog workflow and make it, make the data happen across and and typically now old school we have many point solutions, many different things and this is a problem. Now we can much faster create a process for it for you know, for a ticket handling, customer support or something like that. So this is one layer in here which is quite high up, if I would argue, front end, in terms of capturing the workflow. Then what are the other core dimensions now? Because all that data now becomes interesting when you start storing it, modeling it and then do automation, then do analytics on it. So there are more things to this.
Rainer Deutschmann:But why I'm starting with this example is it's such a relevant example because it's immediate value. So at the end of the day, the question is always, I mean, as we invest into those foundations and the use cases, what's the value I'm getting out of it? And so having an ability to serve the customer faster and better is huge value. And that's exactly what's happening in this particular case, where I don't have to just take a ticket and tell the customer oh, I hear you, but I don't know really what's going on. I can tell the customer oh yes, I know there's a certain problem in my systems and you will get notified. As soon as the system is rectified, the issue is rectified. And that is actually true, because once the system issue has been solved, either by human or by the bot, that ticket is closed and the correlating issues that have been complaining about also are notified, so that closed looping is immediate value to the customer.
Rainer Deutschmann:Give you another example which is very relevant for anyone selling stuff. Right, it's the question of whom to sell what, when, to which channel. That's obviously one of the biggest problems that everyone has that sells who doesn't sell things. So for us in Telco, that was, of course, also one of the core use cases that really shows the value of data. Who has certain products today? What is the usage of that product? How am I interacting? Am I typically a customer going to a store? Am I a customer going and buying and shopping online? Do I have the app or not the app? When do I do these certain things? Do I have a family?
Rainer Deutschmann:So a lot of information is there and that information again is obviously the basis for coming up with relevant recommendations. There's also going in the value of the product. Obviously there's also going in the likelihood for certain transactions that will happen. So having such a recommendation engine which then spits out an offer in real time that offers either the agent will tell the customer or customer finds it online, is huge value for the company. And the new thing is now, with a large language model, I can even craft the presentation of that offer for you specifically in your language, not only language in terms of like Swedish, english and so on, but also in the way how you would best convert, so I can make it most relevant for you how I promote the specific offer. So that's how far we are today.
Anders Arpteg:You can personalize the offer in a way that is exactly what the offer presentation.
Rainer Deutschmann:The offer, anyways, is personalized, but also the way how I'm telling you about the offer is different between the two of you, because you may have different ways of how you have interacted before.
Henrik Göthberg:But I think we can if we slow down here a little bit, because you're highlighting now when we're building a data and AI foundation and we're starting with the workflow layer and you're highlighting, you know the value of getting this connected, but the step in between here that maybe not is so obvious to you. But it's like when we are creating that digital workflow, now we have the data, we're creating the data, we're creating the raw material to do all the really cool things. So the ServiceNow idea why you want to start in this dimension for me it's two major things. That is double-webmy value. It's the value here and now for you selling better topic. Then it's the compounding effects value. That's right that we have the fundamental, we have the raw material we need. Otherwise, we don't have the raw material. That's right.
Henrik Göthberg:And I think this is all sometimes missed yeah.
Rainer Deutschmann:That's right. The signals from my customer, the signals from my operation, the signals from my employees, all these signals I need to be able to avail, first of all, absolutely, absolutely yes.
Henrik Göthberg:And then you come into the other fancy stuff. You know, when we talk data and AI, we want to talk the platform, we want to talk storage compute, we want to talk machine learning. But it starts with the flow that gives raw material and then, from a flow, when you have the basics in place, then we need to have other things in place in order to do the combine stuff and find the insights in it yeah but but you went on a fairly big journey here with the with aws I don't think that's a secret and you, you went.
Henrik Göthberg:Yeah, there were some interesting technology choices here. You serverless quite early and stuff like that.
Rainer Deutschmann:Correct.
Henrik Göthberg:Do you want to elaborate a little bit about some ideas?
Rainer Deutschmann:Yeah, I think and that's one of the P's in those five P's we touched upon is the partner P. I think it's so important to take choices, and those choices, for partners don't necessarily always refer to one is better than the other from a technology perspective, because we can safely say you can build amazing stuff on any of the hyperscalers, you can build amazing stuff with a lot of platforms. I think there's no shortage of features and no shortage of capabilities. I think the point is, I think, on the partnership, who are also the people behind? Who are the people that I'm dealing with and what is the conviction and the willingness to have a joint strategic agenda and a joint incentive? I think this is beyond features and functions and I think in our journey, whether it was at DT, whether it was at Reliance Geo, at Axiata or now Intellia, we have found always strong, selected strategic partnership to be really a crucial factor for success Not a guarantee, anyways, but a crucial factor for success. And so, that said, we never had any exclusive partner for cloud or AI, and we will not have. I don't think it's a single vendor kind of it's not what you want to do a single vendor, but clearly for certain domains we would have taken certain choices just to be also focused in the sense of simplicity, because if you work with 10 different partners, none of them will be strategic.
Rainer Deutschmann:So in the case of analytics, we indeed have developed a certain spike with AWS again not exclusive and why this also has been a bit important is because obviously bringing an organization, in terms of the skills that people pay up to speed requires training.
Rainer Deutschmann:Up to speed requires training. So we have invested significantly into what we call cloud fluency and cloud readiness training across multiple thousands of people, and you would not necessarily do that across all platforms. So you want to have a bit of a focus. So we chose that one partner, aws, and to drive the training and to make sure that we have a certain center of gravity, also in terms of the capabilities in the organization, and that paid off very well. So I think that's the and similar story you could tell. I could tell about system integrator partners, where also there's so many good system integrators in the world, but also here it's about, you know, choosing a partner or a set of few partners specifically for the respective domains, like analytics could be a domain that you know, you work with a bit more closer and you have the right people coming together doing great things.
Henrik Göthberg:But I think then you are not taking this into. When we're building a data and AI foundation, it goes. It can easily be misunderstood that this is a technology foundation. No, no, no, it's so much more. You highlighted now in this foundational build up is a huge investment in training. It's a huge upgrading in competencies. Yes, can we elaborate a little bit more on what's in the foundational parts? I mean like we can talk about processes, governance, practices, partnerships. So how do we understand data and AI foundation beyond the tech?
Rainer Deutschmann:We talked about a few elements the data, the digitization, then also connecting end to end. We talked about the training part. I think the other piece that I found very important simple things like visualization and sandboxes. I mean data. I mean people at least need to understand data mostly. I mean, if you look at actual, I mean we look at data, we look at graphs, we look at colors, we look at time developments, time series and so on.
Rainer Deutschmann:So visualization at scale, whereby the organization doesn't have to resort to an IT person to please give me some data and doesn't have to open up an Excel sheet to understand some data, but basically fires up, as you would expect, just a, you know a browser and you look at whatever you want to look at and you can modify and change.
Rainer Deutschmann:This is the way how organizations should also I call this also part of the foundation is basically democratic. I call it democratization of data access. Nobody should have preparatory you know ownership of data, of course, within the given confidentiality and everything around which is part of the foundation as well, by the way, which is how to really have the right owner on the right data. But I think that's part of the foundation that we have seen, and then on top of that the sandboxing that I mentioned. I mean, I want to then start to be able to have people play with data beyond just visualization, and that requires a certain amount of training, but now, with the tools that we have, it is not really that difficult anymore to get some certain scripting or get some certain data modeling or even learning in place which I can then experiment and later scale out if I get the right result. That also is part of the foundation that I consider to be quite important.
Anders Arpteg:And perhaps with LLMs, you can even have even more self-service analytics in place, where you even may not need to know the details of SQLs and Python.
Rainer Deutschmann:You're hitting so much. I think that was, I think, the most recent projects in Anthropic is such a good example where, of course I mean earlier we would have to build a rack ecosystem where you put in the data and then you can query through the LLM. That still is, of course, the case, but I think we will see increasingly direct ability to interact and have a co-pilot on all my data, and that is Is are being written in PowerPoint and PDF and they're sitting in some database and you do some normal search across all this huge content. The way how all the consultancies are now working is all of this content is basically available through a reg kind of an access and you can interact with this.
Rainer Deutschmann:What is not yet done, I think, is the information access in organizations I think, at least not at scale whereby also I will be able to just work with the data fluently through a conversation that I have in the organization and of course, you can combine those two, which is the consulting part and the information in organizations. Currently there's data queries and you are shifting over big tables and Excel and databases. That will all be gone very soon. So the fluidity and the ability, like I said before, to have less people required for these kinds of strategic and executional exercises will be dramatic, so I think that's you're right. So long story short. The LLM based access to data will revolutionize the way how we can work much faster with a much broader database.
Anders Arpteg:I think you know you take the term foundation. You can think a lot about what that means. I know if you take a foundational model, it usually means it's not that model itself that does it. It's a foundation for other models that do stuff. If you take an AI or data foundation, as we spoke about, it's not really the foundation itself that do it. It actually provides self-service to other people to work with data. So if you have the right foundation, it means that all people can start actually working with data and AI in some way?
Rainer Deutschmann:Yes.
Anders Arpteg:So it's basically scaling up. I guess one of the S's there to the data and AI that you do have if you have the right foundation to do so?
Henrik Göthberg:Yes, because if we explore it from your angle now, anders, we are wanting to scale up, we want to enable the foundation is there to enable this to exponentially scale and go faster. So then you can argue what is that?
Henrik Göthberg:foundation is in relation to serving the speed and innovation and scaling of the others. So, from that angle, what is the foundational piece in terms of leadership? So now we have leadership, we have leaders, we are leaders in this way. We really want to go data. We understand the digital foundation technically, but if we can't use it, if we can't drive the formula one race car, so what is the leadership angle on foundation or transformational leadership?
Rainer Deutschmann:even so, I think the earlier point of curiosity and leaders to be especially focusing on hard problems is very applicable here as well, and I think what I've seen there's also a certain risk, adversity and shyness sometimes in organizations and specifically in companies which have such a big responsibility, like telcos or banks or other like utilities. So we are typically really very careful, for good reasons. So I think leaders are there also to identify and ring fence domains within which we can start to experiment and apply AI technology before necessarily you would scale it out to the entire organization. Give you one specific example when we had, in the old days, chatbots right, we would all be very excited and launch chatbots to our customers and guess what? The call center volume even increased because the customer was just so fed up with the chatbots that they would rather, you know, be handed over to the call center agent multiple times and it was a mess. So so that, so that created a lot of, you know, let's say, risk adversity with chatbots.
Rainer Deutschmann:Now, of course, now with the kind of the conversation abilities that we now find in the llms, not only on text but also now increasingly audio and even video, lip-syncing, everything is now included, different languages, I mean everything in real time.
Rainer Deutschmann:It's amazing, right, as you know.
Rainer Deutschmann:So now is the time to you know, rethink and experiment in a risk, you know, contained way, contained, right what can I do?
Rainer Deutschmann:And then scale from there and I think that is what the leaders are doing good leaders should be doing and are doing to say, okay, let me take that ring-fenced, sandboxed approach, apply the best of technology In the case of a chatbot.
Rainer Deutschmann:I mean, why don't we do a video bot and then get some customers if they want, they can opt in, or we do some certain sub-segment and get them exposed and learn from this, and then there will be issues discovered, but there will, in the end of the day, be an amazing potential, because then, obviously, you will be serving customers so much better by having all product data, all operational data, everything about the customer is available to that bot, and so the customer never has to repeat oh, I called last time, oh, I have bought this one, and let me look up some product information. It will be all much more fluent and the customer service will be dramatically better. So this, I would say, is an example of what a good leader should be doing Open up that domain, create the safe environment to experiment and then, when it works, scale it into a wider use case.
Henrik Göthberg:And that can lead into an even knock on effects on how to understand the foundation then, because that, of course, then means we need to have a corporate governance that is aligned with the containerized experimentation. It's not shying away, but it's not going crazy either. Exactly so, then something what happens with the governance piece here?
Rainer Deutschmann:So here I think it's important to have the capability, once I identify certain use case domains, like the one in the call center, which I think is very applicable especially for LLM-based interactions to identify those and then create those safe pockets where I can experiment. I think that's and it's very possible. I mean, like I said, we can identify sub-segments or friendly users. I mean we can always identify areas where we start and that will encourage people to really drive the experimentation and test out things.
Henrik Göthberg:But, if I'm a little bit, my experience from Vattenfall and then we have nuclear on one hand side, in our corporate letters, in our corporate governance systems, typically we've had very robust governance for things that are not about experimentation or innovation but rather about quite fixed static. You know, for good reason, right? So do we have gaps or do we have things that we need to add typically to our governance? And and maybe this is more a topic if you're a really, really large company that has been operating for 100 years, I've been put scania and vattenfall and and then sometimes there, to me it's almost like the new things we want to do, are people daring to do them allowed to do them, or are they sort of also needing to shape the policies around experimentation at the same time?
Rainer Deutschmann:I think we don't have a fundamental problem so far. I see, even in telecommunication industry I think there is. Of course, in Europe we have rather too many than too few regulations and guidelines and so on. So I think it's probably not a lack of those and so on. So I think it's probably not a lack of those that is the issue?
Henrik Göthberg:Maybe they're set in a way that they are not really people feel scared to get going on it. Yeah, I don't know.
Rainer Deutschmann:I don't think that's really the issue. It is important, I mean practically speaking. When we talk about analytics, ai, specifically when it comes to customer data, of course we always have I mean things like a specific check before we would implement, before we would touch the customer data. I mean those things are quite naturally done. There's specific sensitivity around, of course, the actual payload I mean the traffic data in telecommunication providers and so on, but this has been clear for so many, so many years, so I don't think that should also not be used as an excuse to not move forward.
Henrik Göthberg:I can just clearly say that but now we have spent some time talking about the foundation, goran, I think we should do an AI news segment before we move into the next topic after that.
Anders Arpteg:It's time for AI News brought to you by AI SW Podcast. There is like a small song playing here for your information.
Henrik Göthberg:You can't hear it, but it's fine.
Anders Arpteg:You don't need to put it on.
Henrik Göthberg:Oh yeah, you didn't have. There was the jingle. You missed the jingle. You told me before you told me before you missed the jingle so, but you want to set it up, like you always do, under a single bell.
Anders Arpteg:You know, we usually have a small break of the highlights of recent week's news topics, of course related to AI in some ways and just share some personal favorite topic that has been brought up and just speak a few minutes about it. We try to keep it short but we usually fail. But perhaps we can try to keep it to like 15 minutes max.
Henrik Göthberg:Let's really do 15 minutes max. We have so much more to talk about. And, rainer, if you want, have you reflected anything that caught your eye in the media and the news that has sort of a tech or AI or data flavor?
Rainer Deutschmann:I think the last two weeks have been actually less busy compared to the last update OpenAI has brought out. I mean, obviously, everyone was really excited about the reasoning now being generally available. But I think the one thing and we briefly discussed it, hendrik has been, of course, elon Musk coming out and with a lot of expectation on his he called it party right. I mean the kind of robot taxi presentation, and then also the robots themselves, the humanoids, and so I mean how to reflect it. We know he is typically more on the optimistic side when it comes to timelines. He was talking about 2026, latest 2027, in terms of the robotaxis to be available. So we can discount I don't know how many years, but at least I personally think this is absolutely the right way, how it should be. I mean, we should have vehicles that don't require a driver. I think we are just too weak of a system to operate such machines. It's much better to put this in the hand of a smart system system.
Rainer Deutschmann:I had actually a discussion at the industrial summit earlier this week with some people who were arguing that self-driving will take longer than expected because of the fact that there are these judgment calls that people will have to take, which machines can't. I'm not on that side at all. I think this is very clear that the safety aspect of self-driving cars is much better as compared to a human's drive. I'm very convinced about that. The only thing that may hold it back, I think, is again regulation. So what are the compliance and regulations that are to be compliant? I mean the regulations that are to be complied. I mean the regulations that are to be complied to, whereby I think Europe may be at risk to just again overdo it and therefore we will be on the long end of the launch for such self-driving cars as compared to, you know, us, california and so on.
Rainer Deutschmann:So on the other side, as also we pre-discussed briefly, I mean the stock market reacted very negatively, so the share went down. I was at 7% or something like this, 8%. So underwhelming level of detail was the feedback. But hey, I mean, I think this, I'm not so much worried about this. I think he's on the right track and I think we will see self-driving cars, and I think the the business model to say, I invest thirty thousand dollars, which is the price he has given, I can use that car for myself and once I'm at work, I actually can.
Rainer Deutschmann:Let it, let it go and work, let it work. For me is is, I think, pretty obviously cool and, and I think, will work, so. So then you don't need huge asset owners that are buying huge fleets of cars. It's basically putting the assets at work. That's, by the way, also what telcos and that's the point of scaling, what we all want. We want to have the assets maximized in terms of the value, so we call it sweaty assets. So why would I have a car standing 95 of the time somewhere? Someone I saw the numbers from the show like he said like five, ten hours a week. We call it sweaty assets, so why would I have a car standing?
Henrik Göthberg:95% of the time somewhere I saw the numbers from the show. Like he said, like five, 10 hours a week. It's ridiculous, ridiculous waste of resources. That is not requesting your assets.
Rainer Deutschmann:It's the opposite, exactly. So I think that was what I found pretty positive.
Henrik Göthberg:Yeah, you, of course you had some angles on this.
Anders Arpteg:Let me just give some more details. You know, I think I thought this was really interesting to see as well. You know, for one, he launched the cyber cab and the robot taxi part. It looks really futuristic, really cool design, without any kind of human driver or no wheels, no pedals. You know, just, you know a place where you have a lounge to travel from point A to B, a lounge to travel from point a to b, and they have actually, you know, they rented out like a big place in holland somewhere where they had like 20 cars driving around, so that was like the movie set, like it was almost like, I mean they did a lot of investment in, in trying to build up some kind of you know area where it looked like a city.
Anders Arpteg:It was fenced off, so it was a closed city, so to speak, but they had cars driving around autonomously and they allowed the audience to go into these cars it was a big, elaborate show. It must be said as well and um, and then they spoke about that, and one of course is they they want to build these kind of cars and put them on a road, and he's very optimistic and he said so himself.
Henrik Göthberg:Of course, we'll see if we can't 2026 or 27 or whenever but you can always tell when elon lies, because he smiles and he tilts a little bit and then he says a number but I think it's impressive.
Anders Arpteg:You know what's done with his fsd, uh, already. And if you take the 12, 12.5 version of fsd it is super impressive. And also we already have, you know, robotaxis driving on the road commercially, like Waymo has it, cruise have had, and you know we've seen them do that in closed settings. But you know we can go into the whole. You know who will win the robotaxi kind of question. If we want to.
Henrik Göthberg:That was actually my angle, if I'm going to put my two cents into this particular conversation.
Anders Arpteg:But let me just finish off the topic here, because I think it's interesting what they spoke about there, For one, of course, the cyber cab they also had like a robo van, or he called it something else.
Rainer Deutschmann:Roboven, roboven, roboven. Nobody knows why he's changing the pronunciation roboven.
Anders Arpteg:It's funny and they looked super cool and they actually put some effort in having really futuristic design.
Henrik Göthberg:Did you see it? Look almost like Iron man face. Yeah, did you make that reference? Yeah?
Anders Arpteg:He is the role model for Iron man right, mr Stark? Yeah, Anyway, you know this opened up so many things. You can think of it both technically, which is super cool, but I think economically it's even more interesting. You know he will have manufacturing capabilities to do this really cheap. It will be potentially much cheaper for the users, because then you perhaps don't need to own a car, you can simply go into it, and he had some specific figures for how, per mile, it will actually drop costs drastically for going from point A to B.
Henrik Göthberg:Are you sweating your capital?
Anders Arpteg:Yeah, that's a good idea, and even if you buy one, it will still be cheaper, because then you can have it. You know, do work for you, so it's still also cheaper to buy one, actually, and yeah, it's super cool. And then he also had the optimus bots, you know the humanoid robot that ran around and they will serve in rings and, uh, doing a lot of fun dancing yeah.
Henrik Göthberg:And that was there. And they complained about something here, right?
Anders Arpteg:Yeah, you could speak to these kind of robots, but apparently it was not fully autonomous and there were some kind of teleoperation or remote control for them and potentially it was humans that actually were behind the discussions for each robot who knows? I think this potentially caused the backlash and also the lack of technical or economical details, as you said, and I think it's a bit sad. I mean, it didn't have to make it look like they were autonomous. I mean, if you simply said, because of safety reasons, we have some humans assisting these, and I think even some person that were in the bar area said okay, we have human human assisted Optimus bots.
Anders Arpteg:Now it's not fully autonomous, everyone would accept it. Now, he didn't say it and it caused, you know, the wrong expectations and then it backfired a bit. Anyway, I think it was a super cool thing and the future is looking really bright and, of course, I agree that we will have full self-driving very soon, or soon at least, and it will be a weird time in 10 years when we look back and have humans driving cars Strange.
Henrik Göthberg:Yeah, let's make this the news segment. We don't need to take so many other news. So do you have one news you want to add, goran, the? I was looking in the feed on the comments and one that caught my eye was the. You know, you know, commenting starts and people are commenting back and forth. Was the argument who's in the lead? Waymo or, or Tesla and someone. Tesla is, you know, completely underwhelming. Waymo sort of done this in production, and then, of course, we've had even guests here who are sort of world class on these topics and it's very clear that there's a very different fundamental architecture model in Waymo and the cost of scaling the Waymo model is not proven yet in terms of how much cost it is to set up a new area, because then you need to go out and map it. So what's the argument? Who do you think is in the head?
Anders Arpteg:I think for one, if you just put it in practice today, waymo actually has a safer ride, I would say, but the reason behind it and who will really scale is a different question. So what Waymo does? They have a super set of sensors, they have lighters and radars and they have videos. They have so much more things. It's really expensive to do and they also rely on high definition maps.
Anders Arpteg:But they haven't mapped out the area, so they can only drive in those areas where they have the maps and that's why it's very localized to a few cities, san Francisco. And it can't really drive in other places, which actually Tesla can. And Tesla also removed all the lidars. They actually use lidars for training, but not in production, so it becomes much cheaper and then much easier to scale.
Henrik Göthberg:But still with FSD, is it anyone who goes neural networks all the way through and has thrown out the instructions, like Tesla? Is anyone else doing that?
Anders Arpteg:Not to my knowledge no, and perhaps we should just elaborate very quickly. We have spoken about this before, but what they did in version 12 of FSD and Tesla was, besides having just neural networks for the perception, going from sensory data to some kind of vector space or some kind of latent space, they also use neural networks for the control If you should accelerate or brake or turn and for the planning or break or turn and for the planning. So normally, otherwise, to my knowledge, self-driving solutions they have more manual rules for saying this is how you do the control and the planning is done in this way. So that's not a machine learning approach for those kind of solutions. So the only part of the whole full self-driving in other solutions is just the perception part, where deep learning is used. But they actually have an end-to-end solution, so using neural networks throughout the whole spectrum perception, planning and control and that's a big thing.
Henrik Göthberg:And it scales, especially when they nail that, when they get that right.
Anders Arpteg:And Waymo. You know they don't even have a manufacturing themselves. They have to rely on Hyundai and other kind of car companies to do the production. They don't have the same kind of automation in the manufacturing plants as they do.
Henrik Göthberg:So, Luis, have you followed this? By the way, the Robotaxi and Waymo approach Not more than what you described.
Rainer Deutschmann:I think the question on neural network full end-to-end is very interesting. Those are still deterministic right, Not probabilistic. The training data. If there are exceptions which have not been part of the training, the neural network is a bit of a risky solution as compared to when I do more deterministic, rules-based things. But I'm not advocating one or the other. I'm just saying that the rare edge case issue is the one that is the only concern because you don't train that that is the only concern, because you don't train that Well.
Anders Arpteg:you know the big thing with Tesla and the big advantage here is they have millions of cars on the road today and if you just take, you know the whole strategy or go-to-market strategy that they have. You know Waymo they pay another car company to build some set of cars. They pay human drivers to drive around, to train the data and have the training data to to build the system tesla. Instead, they build their own cars. They sell it because it's a, it's a semi good car to buy and people buy your car and they drive it around and they make money for manufacturing a car which is actually data collection unit for building a full set of driving in the end, and the people are driving that for free, even paying to drive it and giving away the data instead of having to pay people to drive it, and they can collect from millions of cars throughout the world and actually catch the edge cases, which is really the hard problem to much greater extent than the wayode can.
Henrik Göthberg:So then you get to the topic. Who's ahead? Well, it depends on which lens you take on this topic. All right, I think that was more than 15 minutes, but interesting in the end interesting week and some reflections.
Henrik Göthberg:That was valuable. Jumping straight back into the topic, we are picking apart high performance in the enterprises in the age of AI, and I want to switch gears a little bit and move from the sort of the foundation stuff that we all need to get better at in order to, you know, unlock all this into the you know, to really zoom in a little bit more on value. I mean, we started the whole conversation about that and also what drives you in relation to curiosity, technology, as long as it's applied and creates value. So can we just go here as a topic to really talk about that, and maybe you can start on your own? How do we create value with data and ai? You know what's the examples and what is the mindset to have yeah, that's um.
Rainer Deutschmann:One of the things that I think we we can discuss at this point is the the question on business case. When it's about value, I mean, behind a value, typically you would want to have some kind of a business case where there's a certain benefit and there's a certain cost. And I think in what I've seen in organizations, sometimes there's this question of what's the business case for my data foundation, and that's a very tricky question to answer, because the value will not come from just a foundation. The value will come from what I do with it. So I think we need to turn it around and ask what is the value that I generate and what do I need in order to get there? And we talked about a few examples the customer value management, the customer care, the operations and automation. These are three very relevant examples whereby I need to look at those examples, and each of them has a very strong benefit and, of course, they have requirements for building a foundation and specific implementation on top of the foundation requirements for building a foundation and specific implementation on top of the foundation. So for us as organizations, then we need to just think what's my I would say capability that I'm willing to invest and maintain, and what I mean with that is there's a difference between realizing a certain use case as a point solution as compared to realizing a certain use case as part of a much wider ecosystem. I can realize whereby I have platform-based approach. None of them is better than the other a priori, but it depends very much on my capabilities.
Rainer Deutschmann:If I am weak and I don't really believe that I can build own capabilities, own people, own skills and horizontal platforms myself, I will rather buy an off the shelf solution that works for a specific use case. I can buy a chatbot, I can buy automation. I can buy, I can buy certain things that are, I mean, marketing automation, whatever you name it. There's a solution provider out there. Many of them consider to invest into the foundation we talked about so that I can start to build and maintain the solutions much more from my own value creation.
Rainer Deutschmann:And I will use partners and I will certainly use also certain platforms, but much more horizontally than in the first case, where I do point solutions. And so, for example, if I look at the customer interaction domain, if I do point solutions, I may have a specific chatbot for a customer care, but what happens online? What happens if the customer comes to the store? What happens to other use cases where I want to connect marketing? So I will be very confined in my application of that domain, very confined in my application of that domain, and I will again create silos, which is what we talked about before, obviously something we want to typically avoid and therefore my thought, if we have companies that have a stronger technology DNA and they are dependent upon that DNA as a competitive advantage, it's typically better to consider a more horizontal enablement approach and build use cases on top of those enablers.
Henrik Göthberg:So that's a bit of a winded answer to your question, but I think it's quite important that there are two extreme approaches point solution versus a more horizontal enabler-based approach, but let me lean on the conversation we had with another guest here and a storytelling that I use a lot, so it's a little bit about understanding how to discuss the value of the single use case.
Henrik Göthberg:You have like we, then we need to go all the way out into operations and understand how can you increase productivity or how can you have better customer loyalty, or whatever.
Henrik Göthberg:It is Very sharp right. What you are discussing now in two extreme points to me is if we truly believe in the technology, innovation and the exponential productivity gains we will get. The other conversation is you know what happens if we understand how to scale thousands of use cases and we have a compounding effect of use cases that we can't even imagine all at once. So we need so. We are not only thinking about how to build one use case the most cost efficient, but we are thinking about how do we scale that? And when we had this conversation I think it was Emma Storbakke, she said she used that quite successfully in conversations because then you start questioning about how should I then manage data, how should I then do storage? How should I then do governance? Because it's not about just fixing the data. For one use case to work, I need to fix the data in such a way so it's ubiquitous, horizontal.
Rainer Deutschmann:Yeah, that's what I call it Horizontal. Yeah, absolutely.
Henrik Göthberg:So I think this is one of the key thinkings on value Value at scale drives you in one that you need to work on some stuff, versus use the single value.
Rainer Deutschmann:Yeah, the only why there is a consideration.
Rainer Deutschmann:It's not a trivial decision to make whether I go, let's say, vertical, siloized, use case by use case, or I go more horizontal as a kind of an.
Rainer Deutschmann:Instead of enablers I'm building up and the use case is basically more like an application. On top of that, the reason is that, of course, the investment I need to take for a more horizontal approach is a bit bigger. It's not maybe huge, because I can also start small and then scale, but certainly I need to invest a bit more into capabilities. I need to invest maybe a bit more into the foundation. I need to invest a bit more into maybe some of the platforms I need. I may want to host, for example, the Lama in my own, let's say, data center, rather than paying a bit more expensive per token. So there are certain things that I may need to invest into, but then I have a much better scaling law because then, as I scale up my transactions, obviously the cost for doing that will be a lot less as compared to I have somebody where I'm paying a cost per token which then, at the end of the day, will probably kill me.
Anders Arpteg:May I storm on that a bit and just see if we can try to play devil's advocate a bit here. I think we all agree on the data and AI foundation part of it, that it's great to have data for a more horizontal approach that you can see across the systems service now kind of approach. But then the question is, for every kind of different part of the organization, some capability that they have if it's customer support or finance or whatever, or some kind of product that you have they have so much more than data to be able to deliver that capability, and you may have hundreds of different applications used, or thousands in an organization and you can't really change them all in some way, I would argue. I mean, that would be an insane investment. Even so, perhaps you know, is it really to what extent should it be horizontal? I don't think it can be horizontal all the way to the top, to the end user application, or is that what you mean?
Henrik Göthberg:We're talking about different things here, but it's interesting.
Rainer Deutschmann:Yeah, I think you have a point. If I look at a brownfield existing set of disconnected applications relatively disconnected applications that each serve a certain very sharp, sharp use case, clearly we won't have the capacity to rip all of this out or connect all of this at the same time. I think that's clearly. You're totally right.
Rainer Deutschmann:Um, one of the things that we have done as part of the transformation show in the last four years and I've seen earlier has been now to define a clear integration architecture. I would call it through and this is the standardization part of my 3S through a set of standard APIs that I use, especially internally to have systems talk to other systems, and this is a journey. This is not something that you do and then you're finished at some point in time, but it is quite important to identify the core systems I need for certain workflows and to then start to, when I touch the system mostly when I touch the system anyways because of some upgrade to then convert the kind of the maybe point integration that has been done earlier into a much more standardized API integration. I think that is a strategy I would clearly always support and therefore you will not have all the systems connected, but you also have, you see, a gradual set of connected systems across which you can operate more easily.
Anders Arpteg:Let me give an example. I mean, if you have some customer support system and they have some kind of end application, end user application that the agents are working with, in some case they may have some kind of their specialized, let's say, chatbot solution, perhaps it's integrated into the interface, perhaps facing the customers as well, or whatnot. That's potentially very different from other solutions that may be used for finance or in some other product that you have, but they can all potentially feed from the data foundation that they have. So what I'm trying to say is I mean, you can still have a data and AI foundation, even models, llms, as you said that may be used for many different purposes, but there is more than having data and AI. There is the whole application and all the logic and the user interface user experience and you can't throw that out.
Henrik Göthberg:But the way to then balance the point solution or the domain expertise or the domain focus and their socio-technical system over here with the greater good of the enterprise. It becomes part of the foundation. But not necessarily is it that we have a central, monolithic. It's not distributed versus centralized. So we are talking about standardizations in terms of practices. We're talking about standardization in terms of APIs or design patterns. So we can basically choose. Now I'm going to buy a point solution over here, but it needs to have a very, very strong API interface and it needs to have certain characteristics that makes this into a modular mesh of some kind. So I think when we understand horizontal, it's not necessarily horizontal because it's one monolith in the middle, but it's a horizontal thinking, investment and what I would argue then you could almost see the horizontal idea as an innovation tax. I buy a point solution, I buy something over here, but I need to certify or secure the data contracts, the API standards. That makes this interoperable.
Anders Arpteg:But there is a horizontal aspect of it, and that's the data and AI part.
Henrik Göthberg:Yes.
Anders Arpteg:I think so. So there is a horizontal part. I would argue.
Henrik Göthberg:Yeah, but the horizontal part? I mean like to go to horizontal part. Do we put everything in one data lake and we have a monolithic approach and everybody should know all the data, which is not feasible from a bounded context point of view.
Anders Arpteg:I think you need to be able to connect the data from different systems. So in that sense, you need to have a way to interface the data between the systems.
Henrik Göthberg:Yeah, this is true, but what options do we have to get there? You know, we have the data measure principle, we have the data lakehouse principle, but what I'm trying to say there is horizontal solutions, yes, but not throughout the stack.
Anders Arpteg:Right, so there are horizontal approaches, but we need also some point solutions. When it comes to the top layer of the stack, I would argue yeah, yeah, I yeah.
Rainer Deutschmann:I don't think we have a disagreement. I think the issue that I've seen created is when a point solution has a very proprietary and not opened up let's say data and wants to. I see quite a few vendors which, I think wrongly, ask okay, I need just to get all the data feed from you and I do the magic in a black box. And typically I've quite rejected such a solution because that means you're kind of giving away the intelligence and you're confining yourself to that intelligence that you get from that particular solution. So I would much rather like to have the data accessible Again, like we said before, not necessarily the old style data lake, but at least there needs to be an accessibility and the fluidity of the data so that we can have a multiple number of applications using the same data.
Henrik Göthberg:We're on the same page here, but this is a tricky conversation, right? Because, depending a little bit about your history and legacy someone has been in data warehousing, someone has been in ERP Our understanding for how to create the horizontal solution is very different, I find. So we are talking using the same lingo, but people are thinking completely different. So when we are digging deeper, I think we're 100% on the same page.
Rainer Deutschmann:What you said, I think horizontal, I define as able to serve multiple number of use cases. Right, that's really what we talk about. So I don't want to have, just like you said, the customer care being served from a certain transaction data. I want to maybe also have my internal employee. So that's, by the way, interesting as another use case whereby we see, typically, technologies where we help our customers for problems, the same exact technology and much of the same data we can now use to serve our own employees with their problems.
Rainer Deutschmann:Like whether it's IT problems, whether it's even problems in the HR system, whether it's problems in any other of the internal efforts that are being made. So we see more and more if I have it's as a user that has a problem, whether the user is a customer or whether the user is an employee. Actually it's not that much of a difference. It's just different sets of information that are required to serve. The chat interface is actually the same.
Henrik Göthberg:Okay, I think this is a profound topic here that not everybody is following here, because we're talking about value. We need to have a way to understand value from horizontal value creation versus vertical value creation. I use the analogy sometimes how I put the use cases on the Y-axis and have all these different operational business use cases and then I take the scaling or the horizontal use cases, like creating data in AI DNA. So this is the crazy. You know that we are working with value in two different dimensions. At least here Is that a first.
Rainer Deutschmann:No, I don't think so. I think value always comes from the application. Where do I use data AI? Where do I make a difference? Use data AI, where do I make a difference? So, for me, the value only emerges at the point of where the actual usage of a certain use case is being done Serving a customer, serving an employee, automating a workflow, interacting through a video chat and removing the need for having a human. These are the points where the value comes to life, because there I can actually either save cost or I can make more revenue, like we said for selling smarter.
Henrik Göthberg:I can reframe myself to articulate that sometimes, when we work value on the data foundation, we are creating tremendous automation value, but not for the end customer but for another data worker. Yeah, but that's risky.
Rainer Deutschmann:Like I said, that's a risky thing and we didn't probably finish that business case discussion. I don't really think that it's so possible to do value for something that doesn't really deliver a value right? I think in IT generally or in technology, we have enablers and then we have the actual use case, and for a use case I need certain enablers. Now I can buy cheap enablers that are serving very narrowly the use case, which is more the point solution approach we discussed. Or I can use enablers which scale across much more use cases, but the first use case many times has to bear the cost for the horizontal enabler.
Rainer Deutschmann:It's like if I build a car factory, I could also say the whole car factory is priced into the first couple of cars, which obviously you wouldn't do, but it will scale to millions of cars. So same exact issue we have here. So that's why the decision has to be again back to my point. I need to understand what is my willingness to invest and to scale continuously going forward on top of enablers. Am I going to be able to use the enabler at scale, or will I basically buy a factory but I'm only using a room of that factory and I waste the rest of the enabler. That's what I need to understand and that's why I need to decide either a narrow or broader enabler.
Henrik Göthberg:If that is to your point, I think this is very good. I like your point here Sometimes. Okay, One beef I have is that sometimes, when we are looking at sort of the vendors and the technology we are doing, we are talking about very techniques. We're talking about techniques how to manage data, as an example, that has that you know you can use that in order to build capabilities. That ultimately has several steps until it reaches real operational customer value. So I think how to understand value in this sort of value chain or in this you know we are creating value here in relation to the next data worker for him to be more efficient. This is one argument, how you can see it. You don't like that. You want to always understand the capability in order to have thousands of opportunities we have over here, and I think it's very fresh, yeah.
Rainer Deutschmann:I think that's and I'm saying this because we have industries like telecommunication which are very used to such a problem. And what is the problem? Right, we are building a 5G network, spending billions of dollars. I mean, it's unbelievable how much money you need to spend and if you don't have a 5G user, you will basically have a huge cost against almost like no revenue. And then, as you are loading and as you're selling 5G phones, more and more customers will use the 5G network but also don't pay more for the 5G because they're used to whatever data.
Rainer Deutschmann:With 4G or 5G, you oftentimes don't even monetize more.
Rainer Deutschmann:And the trick here is that I have what is called the full cost of the network and I have the incremental cost of network provides an empty capacity, which is extremely expensive, and I'm stupid if I don't monetize, which means if I don't have enough customers using that capacity to the point of spreading the assets.
Rainer Deutschmann:But if I make a choice, should I invest more? I need to look at the incremental cost and that is really the cost that I need to put into for sure. I need to put at the incremental cost and that is really the cost that I need to put into for sure. I need to put into the business case. And that's the same equivalent. If I look at data and analytics foundation, I will need a certain foundation which is a cost I need to swallow, like the 5G license and the 5G base coverage network I need to just have, otherwise I can't operate. So, similarly, I need my data center, I need some certain capabilities, I need some certain platforms that I have, but then if I build a use case, that incremental cost for building that use case, obviously the business case needs to carry off that end application of the end use case.
Henrik Göthberg:And I get your argument.
Rainer Deutschmann:So there's a strategic decision, for that's why I'm coming back to this. There's a strategic decision for me as a management. Am I willing and able to invest into a larger foundational set of infrastructure and capabilities which can help me later to have lesser incremental costs, like the token example we gave? Or am I more cautious and I'm not investing into the platform but I'm investing in the point solutions, which initially is cheaper, but in the long run, if I scale it out, it's going to be more expensive on an operational cost basis. That is the trade-off.
Henrik Göthberg:Yeah, but if you now look hardcore into sorry, this is such a hardcore topic for enterprise and data and AI For how many replatforming investments has been done throughout the years where we build this stuff and we build it and they will come and what you are highlighting now. Be careful now, because there's a slippery slope here that we are investing in tech or replatforming but we haven't really figured out where the revenue will come from. So I see that point. I still don't think that this argument or discussion is solving that yet.
Henrik Göthberg:I think, there is a big pink elephant in the room of how we create ownership and adoption in business to start reinventing their business, to use the data.
Anders Arpteg:Where does that?
Henrik Göthberg:come from? Does it come all the way out from the customer? So it leads to the topic, you know, really push or pull. So how we? You know we think about how we build it and then we try to push it out. But if you go with your argument, it needs to start with pull from the market and then into the business.
Rainer Deutschmann:Yeah, my point, just to maybe conclude, if you want. But my point is, data AI is nothing by itself. The value of data AI only comes through from the use of it, and I can make it expensive or cheap, I mean, but anyways, if I don't use it it's a waste. So I think that's what I try to say, and hence we just need to decide how much do I believe I can scale it out and how much would I want to build some foundational capabilities and invest into them so later I have less operational cost, variable cost. That that's the point I try to make yeah, and I then I buy it.
Henrik Göthberg:I buy that 100 percent uh from an r, but I I see here. You know, this is one of the some of the hardcore challenges. If I look at the industry, if I look Data Innovation Summit been running for 10 years. Have we really moved that much forward in 10 years? To be honest, right, are we working?
Anders Arpteg:on the same problems.
Henrik Göthberg:So it's back and forth a lot, you know. So it depends on which mood I'm in. If I'm positive, we have done fantastic stuff and I'm not so positive mood I'm in.
Anders Arpteg:If I'm positive, we have done fantastic stuff and I'm not so positive. Perhaps that's a good segue to some more future outlook kind of questions.
Henrik Göthberg:I think that's it. Value is hard.
Anders Arpteg:Okay, reiner, how would you think if you were to think of a new company like Telia or something else and you were to rebuild an enterprise like that that is potentially more data and AI driven? How would you think that's potentially different from the organizations we have today?
Rainer Deutschmann:Yeah, so clearly and we had this earlier, I think, fundamentally the way how I think about humans and the role is, we will see us being focused on defining the future, providing and ensuring guardrails to the point of alignment and so on, and handling the hopefully rarer and rarer exception and so on, and handling the hopefully rarer and rarer exception.
Rainer Deutschmann:That's how I look at what really the workforce in a small, medium, large, whatever enterprise will converge to and everything else which is currently still being done manual in disconnected processes and systems will go away sooner or later. Some of the companies that are too late will die, but I mean the successful new ones anyways will build it from scratch in a connected way. The old ones that are transforming will hopefully transform fast enough to not be wiped away and we will see, like I said, the focus on those things the future, the guardrails, as well as, then, the exceptions. That's where we will focus on and that's how the guardrails as well as the exceptions. That's where we will focus on and that's how I would build a company right in terms of the focus.
Anders Arpteg:So, in a more connected way, to have a foundation that is based on data and AI in some way, but that have the capabilities. That is, you know, where humans basically focus on more. The exceptions, perhaps, the more age cases.
Rainer Deutschmann:Age cases and the future right. We need to still have the kind of the vision that we want to realize and we need to, I think, obviously still today. The innovation oftentimes happens at intersection of domains where, of course, in the long run, everything you could argue will be done by the large language models. But I think still there is, for the time being, room and need for humans to be creative and that's, I think, where we will see, for some years to come, at least employment.
Anders Arpteg:Do you agree with a statement, like Sam Moulton said, that yes, we have these foundational models, but there's never been a better time to do innovation, to build a startup, than today?
Rainer Deutschmann:I agree, I 100% agree. So this is back to the start of the very start of today's conversation. This is why I'm so excited also to invest into some of the very start of today's conversation. This is why I'm so excited also to, you know, invest into some of the companies. None of them even dares to think about building a foundational model. I mean, this train is gone. I think the time is now like we had earlier and this is many times being discussed like the inception of the cloud, the inception of the mobile phones. These are platforms which just unlock the next S-curve of value creation. So exactly that we will see, and I think that I'm 100% with that statement. It's a new platform that just unlocks new capabilities and it's about the faster companies that really take that set of capabilities and use it for value.
Anders Arpteg:And do you think enterprises, the incumbents that we have today, will keep up?
Rainer Deutschmann:I think the enterprises, like a telecommunication company, should stay focused, and what I mean with that is the business model of a telco operator is to provide safe, secure, resilient and fast connectivity. That is what we are for Like you have an energy company, like you have a water, and so connectivity that is what we are for. Like you have an energy company, like you have a water, and so on. I mean we should stay and be focused, and we have done many times mistakes to go in all sorts of tangents. That has not worked out. However, the way how we do it is obviously to use the capabilities that are being provided by the new players that we are talking about, and that requires the ability to learn and adopt that we talked about here.
Henrik Göthberg:I think this is a profound statement. We all need to stay focused now because it gets so complex and in order to solve your stack or whatever you're doing, it's just got more complex. So try to do that in many different angles will be impossible. In order to unlock how you do it in the really world-class way, yeah, so the answer is yes.
Rainer Deutschmann:You think they will stand up to startups that may have fast, yeah, okay, yes, I totally see this as a next platform.
Henrik Göthberg:Yeah, but then they need to then rethink. We can take the canvas of what you said as what you would focus on, and then we can take even telco as an example industry. Where does this take us in relation to what is the competencies of the future that we need to get more of or less of? Or how does it flip the balance in the company?
Rainer Deutschmann:That's a very profound question. So what I observe, even with my kids, is that curiosity combined with such a low entry barrier to use those technologies we talk about, it's just amazing, and I think we need to replicate that curiosity and the low entry barrier into organizations so that we can adopt, we can learn and adopt technologies, and that's again to my point. We need to enable, I want to enable people to use technologies for the benefit of organizations, for value creation. This is what we need. So that requires curiosity, openness. It requires, of course, also skill building.
Rainer Deutschmann:Even after one year, let's face it, you are completely outdated. So I think there is a need to be absolutely open for lifelong learning, for training, and which specific area to look into, I don't think we can predict, but I mean the attributes of openness and willingness to learn, willingness to try out in an organization for sure will be attributes that I look at. Maybe to add one more, and I think that is the ability to collaborate. That is one of the other attributes which I guess always have been important, but ever more so the ability to work together and to be able to tell a story, to be storytelling. What is it actually about? What we want to do what is my idea and to be able to communicate the idea in a set of people is going to be another critical factor for our workforce.
Henrik Göthberg:For everything. I mean. You always say, like, what are we looking for in the best data scientist and data engineer? And one of the traits is communication, and the more complex stuff gets, the more VUCA stuff gets. Communication and our way to communicate through the mud is so critical?
Anders Arpteg:Yes, and if we have a neural link in place as well, that will certainly improve.
Henrik Göthberg:Maybe we are running out of time here, but you need to tell us when to break, because we've been on it today and I actually put it in the notes I've been using the word. I've been going down the rabbit holes for many years to try to understand the organizational and management template and theory that we live by. You know we talk about. We learned in school about the economies of scale. We learned of certain stuff supply and demand and I believe some of this stuff is very dogmatically part of our world model. And now we're talking about dynamic adaptability.
Henrik Göthberg:So, we're talking about the world model. I put a name on it before in different articles. We go from economies of scale to economies of learning. We go from efficiency focus to dynamic adaptability, focus to dynamic adaptability focus. Is that something that changes the way we operate or organize or steer companies? So, if you're an AI-driven enterprise, and so the fundamental world model, organizational template, steering heuristics that we use, are they different? Are they the same? Are they different?
Rainer Deutschmann:I think there's definitely changes. I think there are parts in the value chain that still have economies of scale. I don't think that's going away. That's not going away If you look at the telco industry as an infrastructure-heavy industry, if you look at, of course, the large language model, training, the reason why there's only four or a few players that are able to do that and maintain. But there is definitely now a lowering of the entry barrier to use these technologies and build on top. And that is, I think, the exciting part.
Rainer Deutschmann:We touched upon briefly to say that I don't need my own server farm, I don't need my own application, I don't even need a big education, to be honest, because I can just basically start to try out things and if I don't understand, I will basically have my co-pilot who will explain to me. That's the cool thing. So the the level of ability, the ability to, to immediate, the immediacy I would say to, to be able to use the technology for whatever problem I want to solve, is ever more uh, I mean, it's it's really. There's hardly any reason not to do it, other than my own laziness or other than the lack of my own imagination. What I want to do and I was referring to my kids, like they have.
Rainer Deutschmann:My son is like 12 and he started his YouTube channel by brainstorming with Chet cpt about the content and kind of the short form video he could create. Then, of course, creating the content in uh, in chet cpt with the image function, and then cutting it together with some video online video, free, free tools and then publishing it and immediately within 24 hours after publishing, learning about what's the feedback of the audience Is there traction or is there no traction and learn what actually works and what doesn't work. And, strangely, the weirdest videos have the most traction. I don't know why. I totally do not understand why, but it is-.
Rainer Deutschmann:Humans are strange, humans are strange, very strange, and maybe it's Google's algorithm rather than humans, but we don't know. So what I mean is that we don't have that long-term time it takes to overcome barrier where the training or where the tool setup, and so on. That's the change, while the foundational things that we talked about will still require scale and synergy.
Henrik Göthberg:We hypothesized a little bit. Like you will live in a world where you have an abundance of opportunities, so to navigate the opportunity landscape becomes a fundamental skill or fundamental steering point. Landscape becomes a fundamental skill or fundamental steering point.
Rainer Deutschmann:It's not about it's about navigating opportunity landscape and it's about steering or navigating by thesis or hypothesis, yeah, and having determination I think the the dark side of technology, especially when it comes to behavioral platforms or platforms that use behavior, is that you are just sucked into an addictive behavior with the social media that we all talk about, and that's the dark side. So, having a conviction of what problem am I really trying to solve I can do it if I just I'm convicted enough, because, again, no entry hurdle. But I think the risk is that we just don't have enough people that really have a conviction and a determination to solve hard problems.
Anders Arpteg:Do you want to end with a standard question? I?
Henrik Göthberg:think we always have a standard question we want to end with and I leave that happily to you, anders, to set it up, because I tried to set it up a couple of times. He does it better.
Anders Arpteg:Okay, well, we've. We've, I think, already touched some of the questions, but just assume and I guess you do believe that agi will come up at some point.
Anders Arpteg:If you define agi as something that's better than average co-worker for any kind of task in some way, what do you think the future will look like when it happens? And then you can think about, you know, a spectrum of different outcomes. One outcome could be very dystopian. You know we have the matrix and the terminator and the machines trying to kill us all in some way. And the other extreme is a utopian one, where we have AI working together with us and they are helping us cure cancer and fix the energy problem. We have fusion energy. They fix the climate change problem or whatnot, and we live in this world that some people call the world of abundance, where the cost of price and goods basically go to zero and we can work if we want to, but we don't need to, and that could be potentially a utopia. Where do you think we will end up?
Rainer Deutschmann:And you mix into this one the discussion we had last evening when we talk about longevity, just to add one component, which is what happens if you reach escape velocity. Now, escape velocity obviously can mean we go to moon and mars, but here escape velocity means we prolong life, our own life, so long as the prolongation through technology is basically faster than the age. Now you can measure the aging. Uh, each year there's a certain way of how you can do this with your blood values, and that can be above one. That means you age faster than actually you should, or it can be below one. The record is, I think, around 0.65, something like this, years per year.
Rainer Deutschmann:So what happens if we reach, or individual people reach, escape velocity, which means that all of a sudden, rather than dying later and later, you actually don't die at all because you can just prolong longer than the age you add to your life and we discussed that in that session yesterday as well because then all of a sudden you know nobody's dying anymore, but of course, course still, people hopefully get born. So what does it mean economically, societally, and so on and do you want to live forever?
Rainer Deutschmann:exactly, and of course that depends on the quality of life and so on. So if you add that to the question as well, because that is somewhat coming in maybe the same time frame, I would guess so do you think that will happen by the that we could, if we want to, can we solve that?
Henrik Göthberg:problem. I mean like now we're having the AGI problem and the longevity. I think so. Do you think it can be solved?
Rainer Deutschmann:I think so, yes, absolutely, I think so. That is absolutely solvable, but I don't know the consequences. And I think to your question you're asking, my belief is what you're asking, or?
Henrik Göthberg:Yes, we are speculating, yeah.
Rainer Deutschmann:How you tackle. I have a more optimistic view, typically on the impact of technology. I think we are playful and try out many things, but we also understand consequences. We listen to Hinton more and more, now that he has won the Nobel Prize especially, and I think the only thing is, of course, what is the power of not over, but also not not under regulation? I think that's a little bit a critical piece, similar to what we have on the environmental discussion, which is yet another discussion.
Rainer Deutschmann:Do we now just dial down too much on carbon emission, the tax for it and the investments going into sustainability, which is happening right now? More and more investors are going out because it seems to be less required. That's a big risk. So I think on that one, like we have for the AI, do we have enough focus on the risks, I think from also a regulation perspective? Because, unfortunately, if you just leave it to the open market, there's less incentive to really do the right things necessarily. So that's probably the only thing I would say there needs to be the right balance, not too much, obviously, but also not too little. But otherwise I have an optimistic view. What exactly that looks like in the future? I really I mean, don't want to speculate too much.
Anders Arpteg:Plus 50% optimistic.
Rainer Deutschmann:Yeah, but I'm generally optimistic and I look forward to a world where we just don't have the need for driving long distance, because why should I? We don't have the need for having people that are doing the whole day the same work, or manual tasks or mental tasks that are just repetitive. I think I look forward to a world where we can focus on more creative things and define the future that we want.
Henrik Göthberg:And solving the bigger problems and bigger problems. Focus on the future topics Exactly. That's what you said actually about what enterprise is also about. I like that future. So with that, I think we can never seem to do this in less than two hours. I promised you one and a half. I couldn't. Super good conversation, so fun conversation thank you for having me sorry.
Rainer Deutschmann:You tried to do some controversy.
Anders Arpteg:We tried but I'm looking forward to the after camera discussions. I hope you don't need to rush. We understand if you need to.
Henrik Göthberg:But now the fun starts yeah, neuromorphic computing and quantum computing let's go there good stuff.
Rainer Deutschmann:Thank you so much thank you very much for having me.
