And thank you so much for doing this. Last time we spoke was in Davos and in the last few months I think SAP stocks have lost something like a trillion dollars in market cap all because of vibe coding. At the same time you see these disasters where vibe coding is actually leading to outages at AWS, data breaches at Meta. I'm just wondering if you can help me kind of square these two contrasting realities in the market right now.
Vibe coding, all the code generation tools, they will definitely speed up the way how you can code software. Actually, there will be of course now new companies, new startups entering the category with AI native solutions. The good piece is all of the SaaS players, including SAP, can use those tools as well. Now the second piece is even more important: who is the winner in the agentic AI category? Because all software as a service players now say, 'Hey, look at my digital co-worker. This co-worker is the smartest.' And I feel at SAP we can contextualize the large language models with a lot of business process knowledge, data knowledge. And to your second part of the question, IT and governance. Because these agents also need to be governed in a way that they understand which data can I access, how can I share data. So it's actually very, very important that we also have a strong governance layer, talking about cyber, data privacy, and of course also the authorization about what these agents can do and shouldn't do.
The other thing that's really blown up since last time we talked is this open claw ecosystem. It's the fastest growing open source software in history. They're just unleashing hundreds, thousands of AI agents into their email, all their personal documents. And it's actually very effective, but also very risky. And I just saw, I was just at Nvidia GTC, where they announced their framework to try to basically corral that sort of open claw power with their own governance structure over that. Do you think that's going to work, or is there just a fundamental problem with just like unleashing that many AI agents into an environment?
I mean, you have to be definitely careful. I also know customers of us who actually build agents with an open source tool. But I also saw many customers actually who said, 'Oh my god, this agent lost a little bit control because this agent shared data which you should never share within the organization.' And you mentioned some of these enterprise AI platforms now coming up. That's why I'm actually also so confident when it comes to SAP because now, of course, a lot of these partners are coming to us and say, 'Hey, SAP, you have the authorization profiles. You have the business process context what these agents need. Can we not work together to give our customers not only very intelligent agents, but also well-governed agents?'
Your customers, obviously a lot of them are very established companies with huge amounts of revenue, large numbers of employees. They're going to be more risk-averse than a two- or three-person startup, obviously. Do the incumbents have anything to worry about just in the ability to move fast?
Look, when you see how the development of generative AI and all of these modules, speed is now really of high importance. But in the same way, actually we have, of course, a lot of customers, not only the largest ones in the world, but also some mid-size customers. They clearly have governance requirements. They clearly have some auditors who want to still audit their financial results. While speed is of utmost importance when we are building agents, as important is also to have a strong governance layer so that you can extend those agents with all kinds of data privacy requirements because sometimes in industry by industry they differ. Some companies in the banking, they have higher standards on data privacy than others in other industries. So that's why we really need to make sure that we do both because some of these agents will never come to adoption if they're only delivering high value but have zero governance embedded.
Do you see any of your customers actually kind of adjusting their risk profile because they do need to just move so fast right now?
Definitely, because there is a lot of hype of AI and obviously we see that customers obviously want to move faster than what we have seen in the past. They just see the potential, rightfully. But at the end, you can do a little demo and customers get fascinated, but then the point when you put the agent into production, I mean, then the fun stops. So then the agent needs to work. You wouldn't see that a CIO or CEO or CFO risking their jobs because the agentic AI layer in the company goes out of control. So I guess you can do very fast, don't care so much about governance, but to the point when you can excite people that wow, this agent is doing magic things, but then at the time of production, of course, the reality needs to kick in because you still need to run a compliant business.
Let's talk about productivity and jobs for a second. I mean, lots of kids coming out of college today are saying the environment's really bad. They're having a hard time finding a job. AI gets a lot of the blame for that. Do you think AI is to blame? Is that what's going on?
I would say yes. I mean, in certain categories, look, the large language models, where are they strong right now? They are super strong on the unstructured content. They are super smart on publicly available data content. So, it's no wonder, no surprise, actually, that they are already disrupting software development. Because code, a lot of documentation is available, a lot of code is available. You can train those modules already today on that. Same is actually true for marketing. I know a lot of customers, also including SAP, who actually saying, 'Hey, a lot of the campaigns, a lot of the content, I can create with AI.' And you see already the impact of AI on the job market, especially on those functions. Now, I would say, I'm now not so negative to say, 'Oh, these companies will have suddenly 30% less people.' I would just say, you see that certain jobs are going away, but at the same time, and I see this in SAP, you need different job profiles, because there is so much still to develop in the agentic AI layer, that there is not a lack of work, but again, the profiles will look very different, also when it comes, for example, to software development.
