#92, Bernd Greifeneder, Co-founder and CTO of Dynatrace: The Rise of the Chief AI Officer

0:00Causal AI is deterministic and won't hallucinate. But all the fears come from hallucinations of AI. And so suddenly we are faced off getting questions. Oh, is now causal AI as dangerous as generative AI? No, it's not.

0:24We have had several cybersecurity experts on recently, but none more prominent than our current guest, who has an amazing story to tell. Alongside the original co-founders, one of whom happened to be his wife, Bernd Greifenader developed a groundbreaking product based on observability of applications and infrastructure, allowing organizations to trace security threats in real time. This innovation and the quality by which it was delivered launched DinoTrace as a significant player in the unified observability and security space. We discuss pitching for investment in the early days and why the original investors really believed in this product and helped push it forward.

1:06We also delve into how Burnt and the team retained the culture of the organization from the entrepreneurial stage through to the other stages of scale over the last 15 years. We also explore why they chose to set up in the US. We examined their hypermodal AI platform and also why Burnt thinks it's crucial for organizations to consider hiring a chief AI officer in the coming years. This was an incredible conversation. I never tire of speaking to entrepreneurs and business leaders who are this passionate about their field of expertise. And Burnt certainly ticks that box. You'll enjoy this one. It's Burnt Greifenader.

1:57Oh, Bernd, lovely to see you and thank you so much for agreeing to do this. Really excited to talk to you. There's lots and lots of things to discuss today. But let's start off with a question we always start with. What does good leadership mean to you, Bernd? To me, when I get this question, I start off with one word, and this is the word of proactive. This is always the biggest one. Also, this is how I differentiate managers from leaders. This is not by the number of people you have, but it is how much they actually see challenges and be proactive about tackling them in innovative ways and driving change and driving forward versus just keeping the status quo working.

2:40Wow, that's a very progressive answer. I think, do you find that many business leaders do tend to blur the lines between management and leadership? What would you think is the biggest difference and what separates those two things? Yeah, this has come through in all those discussions around, do we have here very technical leaders who have so much value for the company where we actually do not want them to have many direct reports. But yet at the same time, if you build a product and your entire company is based on a product, then you need those leaders even as individuals. So in giving the individual leaders also a proper tiling and career that matches up to the common thinking of, oh, managers, the more people you have, the more career you make, This sort of was also a big discussion in our growth area to make it clear that every individual leader has the same career chance and the same compensation chance and opportunity as a manager.

3:49Because at the end of the day, I could almost argue that the leaders, even regardless of individual leader or sort of broader leader with reports, are more important than just the managers. But eventually we need both of them. No, absolutely. Yeah, absolutely. You hit the nail on the head there, Bern. I completely agree. So for the listeners who are not familiar with yourself, they probably would have heard of Dynatrace, but if they're not familiar with yourself, could you give us a little bit of an introduction on maybe how you got into the technology space and then subsequently the years leading up to setting up Dynatrace in 2011, I believe?

4:25So today I'm the chief technology officer for Dynatrace. And the story I got into this actually started much earlier in the dot-com boom time. There was chief architect for a load testing software. And I had back at that time a customer throwing me the manual in front of my feet and saying, I want this red line to go away. obviously very inconvenient moment for me because the customers stated we have a great product that does the load testing but still they could not go live with their e-commerce system because the e-commerce system would fail already with 50 users trying to go online so but sort of we told them that information super valuable but could not help them to figure out in the insights what was the root cause of that issue.

5:18And this eventually then drove me to test different other products out there and think about how can you analyze the inside of software and tried out profilers that gave you probability trees. But as an architect, I could not make any sense out of this. How can I then fix something? So this led them to the idea of what if you would start with an end-user's click, for instance, on an online shop or think of a banking app, how that one click goes actually through the system from one software hub to the next software piece to the next software all the way to the databases and understand this as a trace.

6:01Eventually, this idea of tracing through applications was the foundation for our company. This is why we are called Dynatrace for dynamic tracing of such user interactions all the way in complicated and complex software systems. So talk us through those years in terms of setting up the company. Obviously, we can see what problem you were trying to solve and how you went with that name. How did it all come about? How did you meet your co-founders and what were the early years of Dynatrace like? Yeah, so the pain of looking into software and kind of helping this started with this load testing, but it was also very interesting that at that point in time, I created for this company another product that was more in the area of real user experience, but they could not sell it properly, or they sold it to the wrong audience, let's put it this way.

6:54They sold a production product to their testing audience, totally disjoint. And this made me then think, okay, I don't want to invent any product near for this formal company. If they don't know how to sell it, this is a waste of my time and energy. And this was for me the trigger as an actually very loyal employee to say, no, I can't do this anymore. I have to do something on my own. And then there was the moment about, okay, I need to start a company. How do I go about this? At least with this background of having been the chief software architect for a NASDAQ-based company with this load testing offering before, I knew I need to build something that's global and not just a local.

7:43And I also had the sense, okay, the software that being built is for enterprises and that costs more in the scale of a car than just a donut sort of comparison. So this required, how do we start there? I'm a technician, so I need someone actually who does then also sales and who does marketing. So my wife then did jump in. She was herself in the e-commerce business, ramping up a business for office supplies, selling B2B. So she knew all those pains the e-commerce shops and their managers had. So eventually she joined them instead, sort of taking the double risk. She was actually even at the seed financing committee pregnant.

8:38So the investor committee said, what? How crazy are you actually? You're a couple and pregnant, so this can't work. But this was a key moment. And I was sort of naive enough at the beginning to think, yeah, I mean, this one million should be good enough sort of to get going. Obviously, after three quarters, the million is gone. So we needed some additional founder helping with finance. And we got then the third one on board who helped then finding venture capital. Yeah, so it helped a lot sort of that we had those clear role deviations of technical marketing and sales and finance. And then in 2007, it was clear we needed venture capital money and eventually then get the first five million on board.

9:37And then also with the venture capital money that comes to request. But you need also a CEO. Especially we took venture capital money in the US. And this is maybe also a lesson learned and one that I say worked out well. initially we wanted VC money local we got actually those partners would not really understand the markets we want to sell to then they always thought this is super high risk what do we do and we had at the end five offers on the table a couple of in Europe but then two in the US and at the end we picked the one in the US not for the money but for the best understanding and believe in the market were proved out as the right decision to go for the one who believed in it.

10:31Because when then later sort of the economic crisis came, we heard of many other startups where then investors bailed out because of risk and they did not understand the market versus other investors completely doubled down on this and helped us even selling and pitched better than our own salespeople. This episode was brought to you by Be Digital. Be Digital support leadership teams to optimize cost and get more out of technology investments. Be Digital and the team have unrivaled expertise with technology license management and data remediation and are therefore perfectly positioned to help prepare organizations for AI technology capability.

11:17And on the last point, Be Digital have just developed a cutting edge AI readiness assessment, which provides tech leaders with a platform they need to make well-informed decisions about AI adoption strategy in 2024 and beyond. Go to bdigitaluk.com to find out more and get in touch.

11:40So what did you do to get that buy-in from the investor you ultimately ended up engaging with? Why did they believe in you? How did you sell it to them? So the investor that we had there and the partner there was the CEO of a monitoring company that itself did an exit of over 500 million. Wow. And this CEO knew from their customers that tracing through a system was to them the holy grail that no one could do. so when he heard the story of dinatrace and he keeps repeating the stories on the first phone call we had he said to me hey bern if this is true i'm flying on monday to lintz to austria and just tell me that this is not a lie because i mean this really so will come and then eventually he came and this was yeah like an immediate understanding from what is the pain sure and what is it that we can provide?

12:46So they immediately understood the value of the product and the product market fit. And, you know, obviously, so that, yeah, exactly. Wow. Okay. Well, I mean, getting that excitement from one of your investors is so important, isn't it? I think, because if they're excited about it, then they'll help push the company forward. They'll be like a salesperson for you, won't they? So, and they're obviously with a very good contact book, I'm sure. So, that's incredible. So, yeah. So, I mean, talk us about scaling then. Obviously, you've been on a journey since that investment right through to today. What milestones stand out for you as being really important moments for the evolution of Dynatrace?

13:23Clearly, it was in 2007 hard for me to hand off the CEO role to bring in a new CEO, but a new, I can't do everything. And this is Definitely something that is an important habit and realization that as you grow, you have to keep handing off responsibilities to other people. And so hiring the CEO was super important. Although the CEO was in the US, so this was a great partnership, me and Europe in the US. We could help work across the pond. And then we moved on, built the team. And the big one was actually to get with a funding of less than 20 million in total positive, which is these days actually not this common.

14:15Usually startups burn much more money till they get positive. And we had then about the run rate of 24 million and got then acquired by another US software vendor. became then part of a bigger ecosystem. And this was sort of then very corporate. Interestingly, they had some other two products to converge. So I was chartered, hey, Bernd, I mean, there are two more products in this field, bring them together. And I figured, no, this is duct taping. I can't create the product or a suite of products that doesn't have a great experience. So I had to convince the management to do a greenfield approach.

15:02This means to actually create a dedicated new brand and a startup within the organization and created a completely new product, sold it to a dedicated target audience. It was all SaaS-based. It was more SMB initially. It was all about innovation. Then I got into the moment of we spent 20 million in R &D, sort of had a great budget, but we did a million in revenue. And then the investor came, yeah, okay, but this does not work, sort of, you're wasting too much money, so the product that is too expensive for this. But eventually we could, with help of comparisons, I mean, I bring up the Kodak story, Kodak invented sort of the digital camera but that then suffered from their fear of cannibalizing their film industry themselves so with this i could convince management no we we actually need to continue on this drive this forward and then the investors eventually then helped move this into then the world of b2b and enterprises we converged it with the rest of the business and then suddenly the hockey stick came in and this drove then the further growth of Dynatrace.

16:19Also the investors then in, was it exactly in 2016, I think, carved out Dynatrace entirely and we became then a standalone company. Again, the whole startup mentality, entrepreneurship kicked in again and we could boost it further leading up to 2019 when we eventually had them Went public then, yeah? Go public, which was a key moment. Yeah, I can imagine. Let's come back to that moment when you went public then. But I'm really keen to hear, Bernd, you were obviously leading the product element of Dynatrace. What was the biggest engineering and product challenge that you ever faced? Yeah, the biggest challenge is actually not technology itself.

17:04The biggest challenge is scaling the teams and making sure that the best people, the smartest people, work together and can innovate properly. And over the course of the years, I learned that with every doubling in size of the team, I have to completely reinvent how we work together. And this is ongoing and this keeps going despite we are far away thousand people in our need. Wow. Okay. So how did you retain that culture that you built in the initial team then? Do you know what I mean? Because I know you needed to reinvent things, but you obviously needed to keep the dna of what initially was the reason for setting up the company the raison d 'etre as they call it how did you retain your culture through the various levels of scale that you achieved between 2008 7 8 right up to 2019 yeah that's a true challenge too because they're still a size in their new team of uh three three five to three hundred people ish all culture is implicit sort of We work together and everyone knows and you just live it.

18:09But then when we grew further and with more people onboarding, suddenly you realize new people don't know what the culture is anymore. And so this was then the moment when we had to switch and make it explicit and initiated from onboarding programs to other programs to make sure we tie them into the organization, set up values and figure out how we help them to live them. Yeah, sure. So tell us about going public, Ben. It's obviously something you put a little thought into yourself and the co-founders, assuming it's something you prepared for for a long time. Was it what you expected? Are you glad you did it?

18:50What's your thoughts on the whole thing now looking back? Having been part of venture capital money, usually what you think is or what this means is you will do an exit to another company. And this is where actually going public was a fantastic opportunity to not just being sold yet to another bigger company and being part of this, but continue the Dynatrace vision and Dynatrace story forward. So I think this was a fantastic shift because for the investor, this was the second IPO they did before this investor had done all these sales to other companies. So I'm very, very blessed and grateful on that, that we could make that move for public.

19:37Let's maybe shift gears a little bit then and talk about the sort of current product set, okay? And the current stuff. So can you explain how Dynatrace's hypermodal AI platform boosts productivity across development, security, etc.? Can you just give us, for the listeners who are not too familiar, can you give us a bit of an overview? So basically, we at Dynatrace, we enable customers to make their software work more flawless and secure. So what this means is, think of the majority of Fortune 500 companies in the world, for instance, use Dynatrace, whether it's a banking app, e-commerce app, logistics app, sort of it's in that world of software.

20:23I think also complexity. Our customers have hundreds of thousands of servers out there, sort of people don't know how those are interconnected anymore. This is too dynamic and too complex. So this is exactly where then this goes beyond humans' abilities to actually understand the complexity for both the reliability of software, but as well as for security of software. This is where Dynatrace's hypermodal AI comes in because first we bring all the data together into one place, keep it and enrich it with context and then apply a causal AI to it that we then complement with generative AI as well as predictive AI.

21:13I can explain a little bit sort of what this means. It'd be good to break down what those three types of AI are for the listeners. That would be really good, just in a very, you know, sort of high-level way. That would be brilliant if you could do that, Bint. Yeah, so because what everyone now has in mind is generative AI, which is the large language model, which is sort of the ChetGPT kind of approach. And when I ask you sort of, what do you want to ride in a car that is driven by CHET GPT? I guess you would answer, no, there's too much. I'd be a bit nervous. Drives you in the river because, you know, it's hallucinating.

21:51You never know what's the outcome. And it's part of actually its strength, but also its weakness is that it's based on probabilistic data. So you never know exactly what the outcome will be, despite very often it's really good. but you never know. So it means this is not really good for an automation. This is also why we demand, this needs human oversight, this generative AI approaches and so forth. What does it mean in the world of software? So when you run your banking software, then you also can't run it the way, like the car that drives potentially into the river. So you need some different approach that is more deterministic, that is more precise.

22:33This is exactly where then the causal AI is very complementary to the generative AI. Sure. Because the causal AI is based on dependency graphs in real time derived from the customer's IT system. Sort of think of all the interconnections of hundreds, thousands of service. Those are captured and modeled in real time. And based on this information, the causal AI understands the causal dependency of an issue to an end user experience or to a security risk. And with this causal understanding, you can provide precise root cause or precise take vector information. So now what we do is we combine that causal AI with generative to provide an easier way for humans then to interact with this.

23:26So you can, for instance, ask, so what was the key reason why last week I had a business drop of 10 % in there? And then this generative UI part of Dynotris Davis' co-pilot would first use generative UI to understand my intent with the question and then ask the causal UI for the details that it knows about the IT system and prepare the answer, provide it back to the generative UI. And the generative UI would then translate this in a human understandable answer. hey, the root cause was actually because there's in this container over there in this cluster, there was this issue and so forth. And this is how you could go forward and remediate.

24:18Sure. So autonomous driving systems, for example, would fall under the category of causal AI, I assume, from what you've explained. Yes. Those are also a mixture of different systems. Yes, they have to be very deterministic systems in order to do autonomous driving. the side of image recognition is then more probabilistic, but this is also why for autonomous car, you have to bring multiple sensors together more than just the cameras. Yeah, that's why it's so difficult to do, I suppose. Exactly, yeah. So I understand you, obviously, I'm sure you're familiar, there's been an emergence of a chief AI officer role in the boardroom over the last couple of years.

25:00What's your thoughts on having someone, And, you know, is AI that significant that we need to have chief AI officers in boardrooms of corporate organizations? Why is it important for companies to invest in something like that? Yeah, right now we have chartered our chief technology strategist in this role, caring about AI. And I think the biggest one is to help with all the misconceptions and misunderstandings that's there. For instance, causal AI is deterministic and won't hallucinate. But all the fears come from hallucinations of AI. And so suddenly we are faced off getting questions. Oh, is now causal AI as dangerous as generative AI?

25:43No, it's not. But this is then to your question of the boardroom. This is what needs to be structured, clarified, bubbled up. And also there needs this help of how to deal with then on one hand, innovation, modernization, but on the other hand, also managing the risks or understanding where there isn't a risk versus where they are and how to deal with those. Yeah, sure, sure, for sure. So what excites you most about the future of AI within the context of Dynatrace then, Bernd? What are you working on right now, which is going to really push your product forward? Yeah, this is the combination of exabytes of data and the combination then also of the different types of AI.

26:31I believe that generative AI alone is a great tool, but is not providing the value that actually everyone is looking for. Sort of we are sort of this area of hype psyche. So the combination with multiple AIs is absolutely key to have sort of more the creative part of generative, but also then the precision part of other types of AI all combined together. And then folded this together with exabytes of data from the systems brings powerful automation and helps customers to protect their software and make it more reliable. Yeah, sure. So exabytes, I think, is a ridiculously large amount of data. One exabyte is, what, a thousand or a hundred thousand terabytes?

27:19Thousand petabytes, yeah. Thousand petabytes. Yeah, the amount of data that's being created is just mind-boggling these days, isn't it? But yeah, so that's really useful. So let's talk about cyber and the cybersecurity landscape a little bit then, Ben. Obviously, there's the emergence of AI, quantum computing seems to be becoming more and more realistic. What are you sort of most concerned about in terms of cybersecurity threats over the next five to 10 years? Yeah, I think one is that you can't protect the system anymore in some form of shielding through firewalls or some form. You have to always think that the intruder is already in your environment.

28:01So you have to go and protect each and every individual piece of software that you have, whether this runs on premises, whether this runs in a cloud service somewhere. You have to protect all those pieces individually. And also live with this that not only hackers will be used on the road, you will, or hackers themselves will use AI more and more to be faster with attempting all the exploits. and also creating and finding new exploits. So this means thinking more about how can you prevent yourselves to withstand zero-day attacks, meaning because you don't have any time anymore, this goes on so quickly, is an important one.

28:50And this is where I believe that you almost like with sort of humans, when you do a vaccination, it needs to be similar with software. you need to immunize software also from the inside because this provides then a shield for each and every little software component on its own. Plus then coupled with the analytics of risk and attack vectors around it, this will emerge. Very other important areas than when we think about how software is being created down the road. I mean, it's complex already today. I mean, too complex, I would argue. then people will now use more and more generated AI to create software.

29:34So alone in the last year, even in those early days, the first statistics say that code redundancy has increased by 30 % already, and we are just at the very beginnings. What this means is code becomes harder and harder to maintain. so as we generate more and more code that is unmaintainable or harder to maintain it means also it is more prone to attacks because all the generated vi does not yet care about whether this is an old pattern new patterns out of it is much harder to actually then secure and protect for instance the generated code with more generations and what's happening but humans do all this so far, we throw more technology on technology.

30:24So you can expect that we throw more AI on helping with also securing and protecting. And this is an almost a recursion. And this is very interesting where this goes. But I also strongly believe with this growing complexity, we will be able to provide help. I was listening to the All In podcast, you know, with Chamath Paliapataa and a couple of his mates in Silicon Valley talking about how the SaaS business model sort of emerged after the creation of the internet, but they think that it will be with the emergence of AI, generative AI creating software, it will massively disrupt the SaaS business model because it'll just, you know, for obvious reasons that you partly just explained.

31:09Did you agree with that? Do you think generative AI will create software to the point that SaaS companies will really struggle to capture more market share. Do you know what I mean? Because companies will be creating their own software to do things. So why would they need to pay license fees for software? No, because you always need to create software that is superior over the average. Yeah. So you think the bar will just raise, they'll just get better and better. Yeah, exactly. Just raise the bar. This has always been this way. Some technology becomes more widely adoptable than everyone uses it.