SCALE OUT

“I think at the end of the day, it's all about adoption and the outcome you bring to the customer. I mean, the technology, look, the reality is for most companies, the technology doesn't matter. I always tell to my developers all the time, our job at SAP is to make the technology disappear. We need to get the outcome in front of the customer. And then try to make sure that we of course, bake the enterprise qualities in and the integration is there. And the customers can turn these capabilities on almost instantaneously.”

“Now we call it agent mining because we record all these decision traces, these contexts, what the users are entering into the system. And then you can either use it to say like, hey, wait a minute, this is actually an anomaly. The folks in, I don't know, in UK from our company or the folks in Australia shouldn't do this because the standard operating procedure is this. Or you say like, oh, that's actually a very good improvement.”

“With AI, the same is happening. It is happening on three levels. It happens, of course, on the UI side. The time is clearly over where you design software, where the dump software that requires the intelligence to sit in front of the computer. Then the second one is, of course, the business processes... a rather rigid process, like the standard operating procedure of a company. But now, of course, with these agents, we can blend the structured and unstructured world more seemingly. And then, of course, below that, you have the whole data layer.”

“Now we call it agent mining because we record all these decision traces, these contexts, what the users are entering into the system. And then you can either use it to say like, hey, wait a minute, this is actually an anomaly. The folks in, I don't know, in UK from our company or the folks in Australia shouldn't do this because the standard operating procedure is this. Or you say like, oh, that's actually a very good improvement. And then you can elevate this to be the new standard operating procedure, maybe not just for Australia, but maybe for the rest of the world.”

“But SAP and these large customers, right? They always have a problem of scale. Okay, what do you know with 100 documents? Well, it becomes a little harder. A thousand documents becomes a deeper engineering challenge. ... we have 20,000 APIs, right? So it becomes just like because it's so huge, right? There's so much things. So it becomes this problem of scale, right?”

“There's the other concept that is explored a lot at NVIDIA, which is this idea of a $0 business. Market creation is a big thing at NVIDIA. Jensen says we are completely happy investing in $0 markets. We don't care if this creates revenue. It's important for us to know about this market. We think it will be important in the future. It can be $0 for a while. An org doesn't have to ruthlessly find revenue very quickly to justify their existence.”

“I would say that the open models are are pretty limited. So, you know, even us when when we try to optimize with our, like, our DGX Sparks, for example, and we try to put them on open models, you know, they're they're getting better, but they're not quite the best. Right? And so just know that when it comes to really strategic work, you're probably gonna be wanting to run on the Frontier models, which are going to cost you money.”

“What hasn't changed is what customers are seeking for, which is outcomes, right? Outcomes and return on their investment in order to get the things done, right? And of course, now AI is an amazing technology that again helps to get more things done in the enterprise, right? And then that is actually what SAP is standing for, right? And so what we are really doing is in given, of course, also the breadth of the portfolio and the customers is, of course, to help customers to achieve more by deeply embedding AI, AI agents, and of course, transforming now the user interface.”

“I mean, for the most part, SAP software is seed-based, licensed today, with a few exceptions. But very clearly, with AI, it was very clear for us that step by step, it will go towards this consumptive world, first consumptive, and then maybe in the next step, once we have more verifiability in the system, then also towards maybe an outcome-based license model. But the reality is also, it is today for us, it's a hybrid model. It's consumptive, but it still has a certain element of seats in there.”

“What we are focusing on is the optimization domains, obviously, and then if you go into things like logistics, traveling salesman problems, knapsack problems, like all these kind of usual hard problems in computer science, these are interesting problems where we believe that could be interesting for the future, for maybe a different kind of computing paradigm to solve for.”

“Now what we do in the past, we call this process mining. Now we call it agent mining because we record all these decision traces, these contexts, what the users are entering into the system. And then you can either use it to say like, hey, wait a minute, this is actually an anomaly... Or you say like, oh, that's actually a very good improvement. And then you can elevate this to be the new standard operating procedure, maybe not just for Australia, but maybe for the rest of the world or more countries to run your company more efficient because now you'll learn something, how the organization behaves.”

“The time is clearly over where you design software, where the dump software that requires the intelligence to sit in front of the computer. If you look at classical software, what did you do? You decide to use an interface. This is over. It's now, we call this Generative UI. The UIs get dynamically generated. If you have analytical questions, for example, or if you want to do your deep research, not just the deep research you find on perplexity or the usual chatbots, but deeply rooted, let's say tariffs are being introduced or new taxes or the straight-o-formals, what does this mean for my supply chain?”

“Large language models are not made for this, right? In a way, how they generate just one token after another essentially in a sequence to sequence modeling, I mean, they're language models, right? And they do this phenomenally well. But if you still want to do these predictors where you have to go back to these classical machine learning approaches... What we said all the time is, okay, look, we have all this data stored in these tables, right? Thousands of tables, right? Where all this information is stored. Can we not apply the same idea that large language models or multimodal models did for the unstructured world, actually for the structured in order to start predicting things?”

“The hypothesis is that, of course, once the hardware matures in the quantum space, there are certain problems that you can address that are hard to address today. What we are focusing on is the optimization domains, obviously, and then if you go into things like logistics, traveling salesman problems, knapsack problems, like all these kind of usual hard problems in computer science, these are interesting problems where we believe that could be interesting for the future, for maybe a different kind of computing paradigm to solve for.”

“My Open Claw and Hermes agents don't do any of that. They run every morning before I wake up. They get better every day, and they cost less than one month of a junior hire. Here's why I would take them over the vast majority of marketers I've ever worked with.”

“Historically, models would be hosted with a single inference engine, and that inference engine would ping pong between two phases. There's pre fill where you're reading the sequence, generating kv cache, and then using that kv cache to generate new tokens, which is called decode. Some brilliant researchers across multiple different papers essentially made the realization that if you separate these two phases, you actually gain some benefits. You don't have to worry about step synchronous scheduling, and you allow yourself to split the work into two different types of pools.”

“Now, the problem is, of course, still today, if we look at these predictive questions, right? ... the challenge is large language models are not made for this, right? In a way, how they generate just one token after another essentially in a sequence to sequence modeling, I mean, they're language models, right? And they do this phenomenally well. But if you still want to do these predictors where you have to go back to these classical machine learning approaches...”

“SAP is the market leader in enterprise, software applications and platforms. It has 400,000 enterprise customers. Usually, I just running their finance, HR, and supply chain, manufacturing, execution, logistics, warehouse management, and then of course everything on the customer side, sales services, commerce, procurement, you name it. End-to-end, like SAP, we always say we have the broadest portfolio in terms of end-to-end running the business end-to-end. This is where SAP started with, giving real-time insight. Usually, I really describe this as it's not just software in itself, it's kind of the operating system of a company essentially.”

“The hypothesis is that, of course, once the hardware matures in the quantum space, there are certain problems that you can address that are hard to address today. What we are focusing on is the optimization domains, obviously, and then if you go into things like logistics, traveling salesman problems, knapsack problems, like all these kind of usual hard problems in computer science, these are interesting problems where we believe that could be interesting for the future, for maybe a different kind of computing paradigm to solve for.”

“The most important thing from a development perspective is actually people start writing their evals. That is, I was on this tour for a very long time because the problem, why does agenda coding work so well, Sarah, is of course, you can verify the outcome, right? You can either say, hey, is the program compiling, or are you unit tests, right?”

“Coding agents have been so much more effective than general purpose agents. And I think a large part of that is it just has access to the terminal, and that means it has access to everything that you've installed into your terminal. It can write code, and it can compile the code. And if there are errors, it can fix it. It can run your suite of tests because that's all just in your terminal. Computing began with a terminal with a shell, but we said that it's not empathetic to humans, so we built these nice user interfaces. And then now we have LLMs navigating our user interfaces, and ironically, we're not empathetic to the machine anymore.”

“We have a couple of people at NVIDIA. We've been working with security to bring agents really close to compute. So we now have stuff where you can tell Dynamo, like, go run some experience with Dynamo on x cluster and just try it right now. We've actually been able to one shot problems. We used to have this problem where, with Dynamo, you have to find the right configurations. We've just had an agent just completely one shot that. It goes. It gets the compute. It runs a couple experiments. It's like, this is the best. Go run this. And then we just give that to people, and it's faster than anything that they have.”

“SAP is the market leader in enterprise, software applications and platforms. It has 400,000 enterprise customers. Usually, I just running their finance, HR, and supply chain, manufacturing, execution, logistics, warehouse management, and then of course everything on the customer side. End-to-end, like SAP, we always say we have the broadest portfolio in terms of end-to-end running the business end-to-end. This is where SAP started with, giving real-time insight. Usually, I really describe this as it's not just software in itself, it's kind of the operating system of a company essentially.”

“Brev was just it's a developer tool that makes it really easy to get a GPU. It was actually Evan Conrad, SF Compute, who was just like, you guys are two dudes in the room. Why are you pretending that you're not? And so then we were like, okay. Let's make the logo of Shaka. We brought surfboards to our booth to GTC, and the energy was great. My wife was, at the time, fiancee, helping me put these vinyl stickers on and she goes, you son of a, if you pull this off. And so, pretty much after the acquisition, I stitched that with the acquisition. I sent it to our family group chat.”

“With AI, the same is happening. It is happening on three levels. It happens, of course, on the UI side. ... Then the second one is, of course, the business processes like an order to cash in the past. ... And then, of course, below that, you have the whole data layer, right? The whole data layer of bringing, of course, SAP has a lot of super valuable data for a company.”

“I have a twenty four seven agent running. I hooked up to run pod. It doesn't shut down instances, and I've tried prompting it. I've given the instructions, shut down when you're done. It's like, I need to keep it warm. I'll need it soon. It's horrible on time estimates too because it's like, yeah. I'll need it in forty five minutes. Forty five minutes of human time is actually three minutes of agent time, so it's like, I'm booting it up. I'm waiting. I'll just leave it on all night.”

“And we believe that still will continue, right? Because this is exactly what we're also seeing right now with, of course, there's still, of course, there's tremendous progress, but we also see that the AI adoption in the enterprise is still not where we want to see it, right? Like there's this Gardner curve, right? Where say like there's this AI innovation race, and then there's this AI outcome race, right? Then the gap almost increases, right? Versus getting narrow.”

“Now I do wanna emphasize this again. Hermes learns from every session. So imagine this. After 20 to 30 tasks in any one domain, it's going to be measurably better than it was before. OpenClaw doesn't do that.”

“Open claw is an autonomous agent, and a lot of people talk about it. And open claw is great for being the execution. Okay. You want it to be your arms. You want Hermes to be your brain. And the cool thing is you could have each of them hold each other in check.”

“There's a summarization of that trend that I like to say to my team. This is the year system as model. Where, instead of having a single model be a thing, you have a system of models and components that are working together to emulate the black box model. So when you make an API call to something that's like a multi agent in the background, it still looks like an API call to a model. Under the hood, it's like a billion different models.”

“Most people don't think about data integrity. They don't think about keeping their CRM clean. They don't think about any of this stuff. Okay. And so you're gonna have to keep that clean. The other thing too, is if you don't have good skill dot MD files, meaning that you don't have good processes within your company and you don't pass them over to, like, your Obsidian to have that central memory continue to get better over time in terms of how you make decisions, well, then this is not gonna be that helpful because garbage in garbage out.”

“But once you have Obsidian added in, it makes it easier to manage everything. So, look, it obviously can't be all sunshine and rainbows. Right? So there are some things there are some limitations here.”

“Agents can do three things. They can access your files. They can access the Internet, and then now they can write custom code and execute it. You should really only let an agent do two of those three things. If you can access your files and you can write custom code, you don't want Internet access because that's one issue. Vulnerability. Right? If you have access to Internet and your file system, you should know the full scope of what that agent's capable of doing. Otherwise, malware can get injected or something that can happen.”

“A lot of people like to say, oh, it's Hermes versus OpenClaw. It's OpenClaw versus Hermes. No. You wanna have both because there's a a chain of accountability that wasn't there previously. And now if you have this on your agent fleet, things are gonna run a lot more smoothly.”

“Kimi two comes out, right. And it's an interesting model. The creators of Kimi k two actually talked about it in a blog post. Attention scales to the number of heads. They made a very specific barter in their architecture. They basically said, hey. What if we give it more experts? So we're gonna use more memory capacity, but we keep the amount of activated experts the same. We increase the experts' sparsity, and we decrease the number of attention heads.”

“For me, it was always very clear. I mean, for the most part, SAP software is seat-based, licensed today, with a few exceptions, like Concur or Fieldglass, for example, or the business network. But very clearly, with AI, it was very clear for us that step by step, it will go towards this consumptive world, first consumptive, and then maybe in the next step, once we have more verifiability in the system, then also towards maybe an outcome-based license model to, for example, what Sierra is doing and so on and so forth.”

“But SAP and these large customers, right? They always have a problem of scale. Okay, what do you know with 100 documents? Well, it becomes a little harder. A thousand documents becomes a deeper engineering challenge. Last year, everybody could build an MCP server. It was so super simple to hook up your MCP server and do amazing things with it. But that becomes like for 10 APIs, not an issue, 100 because you'll get already context bloat and all these challenges. But we have 20,000 APIs, right? So it becomes just like because it's so huge.”

“SOL is actually I think of all the lessons I've learned, that one's definitely my favorite. The speed of light moves at a certain speed. So if light's moving slower, then you know something's in the way. So before trying to layer reality back in of why can't this be delivered at some date, let's just understand the physics. What is the theoretical limit to how fast this can go? And then start to tell me why. Because otherwise, people will start telling you why something can't be done.”

“LLMs, unstructured world, that's all good, right? But most of the time, if you want to plan forward, if you want to make good decisions in a company, you need predictions. Now, the problem is, of course, still today, if we look at these predictive questions... large language models are not made for this, right? In a way, how they generate just one token after another essentially in a sequence to sequence modeling, I mean, they're language models, right? And they do this phenomenally well. But if you still want to do these predictors where you have to go back to these classical machine learning approaches.”

“SAP and these large customers, right? They always have a problem of scale. Okay, what do you know with 100 documents? Well, it becomes a little harder. A thousand documents becomes a deeper engineering challenge. And now if you go into Julia or Sarah, you're maybe an SAP US employee, right? Of course, if you ask a question, of course, for travel policy, for example, of course you expect a very different answer than me as a German employee would get. So you now need to connect this actually with your master data.”

“What hasn't changed is what customers are seeking for, which is outcomes, right? Outcomes and return on their investment in order to get the things done, right? And of course, now AI is an amazing technology that again helps to get more things done in the enterprise, right? And then that is actually what SAP is standing for, right?”