Welcome, everybody. Thank you very much for joining us, and Sebastian, thank you very much for coming here and joining us in the conference this year. Obviously, SAP, we have Mark Moerdler, Richard Nguyen, and Sebastian Steinhaeuser. There is a pigeonhole link, which can be accessed through the QR code over there, and if you have any questions, we'll do them at the end. Raise a hand, and I'll come around with a microphone. Thank you very much.
I'm going to start by reading the Safe Harbor statement. My arm was twisted in order to read. This is not my safe start, however. It's theirs, OK? During this fireside chat, SAP will make forward-looking statements, which are predictions, projections, or other statements about future events. These statements are based on current expectations, forecasts, and assumptions that are subject to risks and uncertainties that could cause actual results and outcome to materially differ. Additional information regarding these risks and uncertainties may be found in SAP's filing with the Securities and Exchange Commission, including but not limited to the risk factors section of SAP's 2024 annual report on Form 20-F.
I think you're a highlight for our next earnings call.
It would be my pleasure. I'll come to Germany. We have the Pigeonhole open. If you'd like to put your questions into Pigeonhole, that's fine. We also will have a mic going around so that questions can be grabbed from the audience. Let me start at a high level, and by the way, I really do want to thank you for having come. I think it's really very much appreciated. Where is SAP on its cloud journey? How much has moved at this point to the cloud? How much will move to the cloud over the next few years? How should we think about that journey and that opportunity?
First of all, Mark, Richard, thank you for having me. Thank you for having SAP. Thank you for joining. So I'm Sebastian Steinhaeuser. I'm the Chief Operating Officer of SAP, and my main responsibility is to drive our strategy to execution. And at the heart, our strategy is how to bring together applications, data, and AI to create a unique flywheel of value for our customers. Now, the cloud transition story. So I actually joined at what we would call the start of the accelerated cloud transition journey with RISE with SAP. And back then, look, many customers had questions about, hey, can I move my most mission-critical system, my ERP, to the cloud with SAP? That's for me the what. Can you operate that system securely in the cloud? Can you do that at a good TCO? Can you provide the right value?
I think those questions I'm rarely getting anymore. I think customers, by and large, have actually accepted that we can run and actually welcome us to run the ERP, particularly in the cloud, for them by now. And actually, AI is a fantastic accelerant to that journey because customers intuitively understand that when SAP develops embedded AI, embedded agents in a finance process or a supply chain process, we, of course, do that in the most modern version of our full stack that we are offering and not in an old legacy version of the software. So I actually see the what is absolutely clear, and AI has even sharpened the case for cloud for our customers. And I think we'll get more into that later. Now, the how, of course, is then something that customers have to decide when is the right moment in the journey.
While many, many customers have gone on the cloud transition journey by now, we still have, and we publicly say that, EUR 11 billion in maintenance left to convert. Then we go on the what I would call the 2X, 3X, 5X journey with those customers. 2X is the initial move from on-premise to cloud, where we take over the infrastructure and the operations of the software. But we don't only do that. I mean, with many, many, we publicly said 50% or more, but in the large transactions, that's even more. Customers actually already at the start opt for also adopting BTP, our business technology platform for extensibility, adopting BDC, our data platform, and adopting AI in the same moment, plus the tools we provide to help them on the migration journey. That's the 3X.
And then there's the 5X because, A, those platforms are consumptive, so there's expansion over time, upsell. And then, of course, there's line of business where we actually are winning market share. So it's a very healthy long-term journey, I think, that is nowhere near ending in the next two or three years.
Excellent. So maybe I'll jump on to the AI. We've been saying that SAP is using AI as a carrot driving cloud adoption rather than thinking of it as an immediate revenue source. Obviously, from what you're saying on the lift, it could be one more recently. Do you agree with that approach or that summary of the way you look at it, that AI is the carrot because people need to get in the cloud before they can be able to adopt the AI capabilities? And why does the cloud transition need to occur first in order for this to be successful?
I don't agree. Argue with me. Not that I don't like carrots. I also don't like sticks. But because for us, fundamentally, AI is not a business. AI is not a carrot. AI is our future business. Now, customers want to adopt the most modern AI solutions. Now, however, what customers do intuitively understand is you don't just want to do what I think in the past has happened several times in IT history, plaster something on top at the UI level. What you want is AI that is deeply embedded into that has the contextual understanding of your business process, of the data that resides in the application, the semantical richness. That's for me, if I think about AI, there has been now a wave of technological progress. The next leap when it comes to AI will be about how to apply AI to business processes.
And that's something where I feel we are very, very well positioned to drive that next wave of AI adoption by really deeply embedding AI into everything we do. So that's why I say AI is not a business for us. AI is not a carrot for us. AI is our future business we are driving. But then AI still needs a runtime. AI still needs parts of the application. So it's not that the application goes away. It's really, in our mind, it's this flywheel of apps, data, and AI coming uniquely together to provide value to our customers. And when I look at, I mean, it's Q4 now. I have many customer conversations. Customers absolutely and intuitively understand that. It's actually even more that AI is not just, of course, accelerating the cloud journey because to do all of that, you need to be in the cloud.
The data needs to reside in the cloud. So it's more a means to an end. It's not that AI is the carrot to move to cloud. AI is the necessary means when we talk to customers these days to adopt our AI embedded in the cloud. And with that, the case for modernization gets stronger. And what customers want to discuss with us is, hey, SAP, what's your agentic roadmap for finance? What's your agentic roadmap for HR? What's your agentic roadmap for procurement? That's where the conversation has shifted. Then, how, SAP, can we add non-SAP data into the mix with Business Data Cloud? We now have a very good answer of how to actually expand the tools available to AI beyond the SAP estate. And then how do I modernize my entire landscape?
Because what customers, many customers intuitively get, it might be simple to quickly code an agent POC on some general purpose platform. But 90%+ of the work is not in the initial POC. It's in operating, maintaining, security, constant improvement, logging. So no customer can do that at mass scale for hundreds of agents. So they much rather look at us, hey, where I bank on SAP, I want you to deliver the full stack. And they put pressure, positive pressure on us to deliver the right agentic tools for them. And I think we are in a very great place now for this next wave of AI adoption with that.
A lot there. Maybe I'll try to unpack a little bit and drill in on a little of what you just said. The approach to agentic there is that an approach where you're finding places of pain and ROI and building focused agents? Or is it more of an approach to give someone a toolkit and say, go have fun and build your own AI depending on what you specifically need? Or is it a combination of that? What's appropriate?
So it's a combination. I mean, of course, our heritage is we are not a horizontal technology or infrastructure company. Our business is of applying technology to business processes. So for the most part, what we are delivering is packaged agents for finance processes, for procurement processes, for recruiting processes that come out of the box and are scalable and take the complete pain away of the customer to operate and maintain their own fleet of agents. However, all of those agents are built on our AI foundation in our business technology platform. So we have rolled out a super strong toolkit that you can now also apply to SAP-centric bespoke industry and customer use cases where I see more and more interest, and especially, I know you're a big fan of supply chain. Supply chain is a very hard area. It's very industry specific.
Only a few tech companies have cracked supply chain at scale, and that's an area where there's a lot of interest because what we uniquely can bring to the table is a super strong, the strongest, I would argue, horizontal platform of any vertical player to build agents, but also the strongest vertical capability of any platform player that puts us in a very unique spot to actually ship custom agents in areas where the customer will opt for custom, but also use the same foundational technology toolkit to actually build very bespoke solutions for customers like SAP has done for 50 years in whatever it might be, last mile delivery in the beverage industry or supply chain spare parts optimization in very specific manufacturing, and that's typically where a lot of the value will accrue.
That's why I feel so confident that both in the standard functions we are serving as well and in those hard industry specific domains, we are very well positioned to capitalize on AI. That will drive the next wave of AI adoption because that's nothing an LLM can learn on the internet. That's our IP in large parts. It's our code base. The agents understand our code. They don't just understand us at the UI level. They understand the data model that sits within SAP. That gives us a lot of great fuel into the AI fire.
Yeah. So you mentioned the supply chain. We talked to your major services partners. And right now, over the past two years, in fact, the area where they invest the most in terms of resources is in the supply chain. So can you share with us about where you see your customers today in terms of the transition to the cloud? And also, one of your major services partners told me something that was really striking: that when they do the benchmark regarding agent AI for different functions, they found that in the supply chain, the return on investment was about 30%, 3, 0. It was the highest ratio that they see across different functionality.
So could you help us understand a little bit better about your supply chain strategy, where your customers are today in terms of the journey, and how AI could accelerate the adoption here with the CHEP2, CHEP3 suppliers?
So I love what you said. I mean, for us, I mean, first of all, of course, there are macroeconomic factors at work that actually force companies or nudge companies to have a very big focus on supply chain optimization right now. And they realize more and more point solution will not help a piece of warehouse software, a piece of planning, a piece of that, a piece of that. But they need integrated because everything is connected. And there we have, I mean, we talked in the past, for those of you who are not aware, we are running the largest business network of connecting suppliers and buyers in the world every day. We connect millions of companies to match supply and demand. Just think of what type of a data source that is to embed into supply chain planning, to embed into manufacturing planning, to embed into.
And then, with that being able to react almost in real time to changes in the demand or supply environment, to react to either supply disruptions or demand changes, shifts, regulation. So I believe that's, and that's a fantastic use case for AI, especially for companies that can combine because it's, as I said, often very bespoke of how you actually solve problems in that industry, sorry, in that vertical of supply chain, core value chain transformations where you need both a super strong platform, but the vertical know-how. And by now, we can bring both of that to the table.
So following up on that, if we use supply chain as an example, you commented on delivering very focused AI capabilities. You also talked about bespoke. Can you give a little more color on how you do both of those effectively and how easy do you make it for the clients? Because one of the issues I hear is everyone talks about they're going to use agentic platforms, and then they go down a rat hole trying to understand how to build them and where the data sources and everything else are.
So I mean, the starting point is, of course, in all the verticals we are active. We are shipping standard software that comes with prepackaged agents that have all the data sources, all the authorization concepts, and everything you need to start operating. But then our, we call it Joule, our agent capabilities, our agents are being shipped with extensibility concepts embedded. We are now putting live in Q4 with GA Joule Agent Studio where you can then actually extend those agents, maybe add a new data source, add an MCP server, add whatever tool you might want so that you can act, because typically that's really where the hard sausage making is, is then how to actually deploy an agent in a large enterprise IT environment where there's a lot of specific needs. And I mean, that's what we've done for 50 years.
We know exactly how to deploy technology in such an environment. And then, of course, that very same toolkit that we use to build our own agents that we harden out on top of the best technology partners out there, the best open source technology out there, we can then also offer that as part of our platform capability to build agents and run agents that are related to SAP-centric use cases. So that gives me a very good feeling about that we can actually handle both. Now, again, we are not a horizontal technology company. The power comes from even when you build agents on our platform, you will have access to things that are unique to the SAP verticals that we are in, data products of our applications, for example, process context that sits in our application.
That gives us, I think, a very neat position in the marketplace where we are not just a general purpose horizontal platform, but also not just a one vertical AI company, but we can actually deliver AI that is opinionated where it can be shipped out of the box, but extensible and customizable to actually work in any industry and any functional environment.
So where are the clients in the adoption of AI? Because we're focused on the S/4HANA Cloud customers, correct? Any sense, any color you can give on how much is being adopted, how much is used, how much is in production, where they are on that journey?
So, look. I mean, what we've publicly said, I think last in Q3 earnings, is more than 50% of our customers are including AI in their transactions with us. Now, remember, there's a lot of volume transactions in there too. So I would say when it comes to strategic transaction, you can think of a higher number. And then we do see very healthy traction. Take SAP as a customer. I mean, we have thousands of users every day using Joule across our estate in finance, in HR. Concur just went live for SAP. I actually will tomorrow show in a meeting how to book with Concur in just seconds with agentic AI your business travel by just chatting with our Joule chatbot. So that's real here and now. And we see real here and now adoption of our AI use cases significantly improving user experience, improving productivity.
And there's a very near-term impact that will have on productivity of our customers that I think we have very, very convincing technology to show here and now, but also a very convincing, especially agent roadmap. I mean, what we talked about in our recent event, SAP Connect, that we will evolve from agents to assistants so that we will take an opinionated, no one wants to operate. I think it doesn't matter if it's 10, 100, or 1,000 agents. That's, I think, the wrong thing. You want something that assists a controller, assists a treasurer, assists a recruiter. You don't care how many agents there are. You care about getting the job done, getting the productivity realized. So that means we are investing into the agent-to-agent orchestration across SAP and non-SAP.
We're investing heavily in areas that are, I think, very unique where we are building very unique API on how to abstract from the model providers so that we can actually route requests to whatever model is best fit for the purpose or the customer needs, maybe, for example, Mistral here in France. We are investing heavily into how to actually make tabular data accessible to AI, and especially LLMs are still not very great at giving 100% accurate answers. They are still, but they cannot hallucinate when you close the books with SAP. So there's a lot of great stuff coming that will deliver tangible productivity impact here and now.
Sebastian, could you please remind us about the monetization strategy that you have and what kind of evolution do you see there?
Yeah. So actually, that's a great topic. I have the fortune of running the commercial strategy for SAP. I'd never tell that when I talk to customers. I always get a lot of feedback. But so for me, AI naturally lends itself to a consumptive metering, and that's what we are doing. So we have an AI unit concept that is basically a consumption pool customer subscribe to because many of the use cases have two properties. A, they cut across several SAP applications, which is a strong case in point for a suite with SAP. I mean, if you want to get something done in order to cash, you're not just in the financial system or the CRMs. You're actually cutting across multiple systems at the same point in time. Second, I mean, there are agents that are attached to a user. There are agents that are machine triggered.
So the consumption meter gives you the full flexibility. However, what we've done in a very opinionated way is where we see agents attached to a specific user, we have introduced that concept of assistance, and with that, a per user meter for finance for HR that is, however, embedded in the consumption entitlement and then allows you to scale up and scale down on a per user basis. So think about it. You have AI with SAP. Maybe in January, you need more HR AI users and agentic calls because you do the goal setting process. But then in April, you close the books, so you need more in finance. So we combine the flexibility of a consumption concept with the predictability of a per user assignment because that's also something, I mean, like it or not, I mean, IT buyers like predictability and the purely consumptive model.
While everyone is always raving and asking for it in theory, in practice, I cannot count the number of customers who then come to me, "Oh, that's all nice, but I want a more predictable model." So we've, I think, found a good optimum how to walk both worlds, retain the monetization flexibility while giving the predictability to the ultimate buyer of the AI. So it's a hybrid. And then, of course, that's an attached play to our solution. So I mean, every transaction that we are doing, we are attaching AI to light up the AI capabilities in every of our line of businesses for our customers.
Following on that, because that's an interesting answer. One of the complaints we always hear from clients, from customers of consumption models is they have no control over the spend and the fact that they turn the product on and suddenly everyone plays with every little tool that's out there and they get ridiculous bills for that. Is this the way you keep that from becoming a pain point?
Exactly. I mean, basically think about it. You have an AI unit pool, and then you say, "Okay, I entitle 10 finance users to use agentic AI from SAP." That comes at a fixed monthly run rate per user, but is metered against the AI unit pool, and you can scale up and scale down. And then there might be other agents. There are other agents, for example, in LeanIX, where user is not the right concept that are not packaged in such a way that are directly metered against the AI credit pool. But that gives the IT buyer a lot of control while actually letting the user do whatever, like go free. So I think it hits a very nice sweet spot between the two.
Makes a lot of sense. It's a good approach.
Business Data Cloud, by the way, is a similar approach, like a consumption subscription hybrid. That's very much welcomed. Enough predictability and enough flexibility, that's a good trade-off.
Going to come back in a moment to the Business Data Cloud. I've argued that SAP has taken a different approach to AI than many of the other companies that I cover in both delivering discrete functionality using a wide range of different models. Most people are going, "I'm a XYZ. I'm an OpenAI shop, and that's the main models we're using." You're using very focused, very specialized models in many cases. You're also running it across multiple different hyperscalers. Why the approach and what does it add to the complexity of the business? What does it add to the value to the clients?
So, I mean, first of all, it's flexibility. It's flexibility for us, but also flexibility to the customer. It also gives us a lot of flexibility in the way, I mean, yes, when it comes to hyperscaler, I mean, many software companies, SaaS companies natively build on one hyperscaler deeply using their services. We have decided we go much more on a, how do you say, abstraction approach where we can use multiple hyperscalers. We can actually, and because that gives us further reach into regions that helps us to cater to customer requirements. And we've taken the same approach for LLMs. We've long discussed, "Oh, should we build an own LLM?" I'm quite happy we didn't.
I'm quite happy we didn't enter the infrastructure race because actually, in a way, that will be the road that's being paved for us to apply the technology to the verticals we are in. So our investment is really into how to be the most flexible in using the best technology for the best use case through our platform capabilities. And then the biggest focus, of course, is in the application of AI.
Yep.
Maybe one thought, of course, what that also means is while we are not a horizontal tech company, the platform investments are absolutely critical. There are only few companies that have our scale to do both, to be a verticalized company, but do platform investments at that scale because over time, customers will want choice of the LLM. Customers will want choice of the hyperscaler. That creates a very high bar to jump over having such and a lot of scale benefits when you think of a best of suite versus best of breed approach.
Perfect. I'm going to follow up with one of the questions from the audience. Can you speak to the effect of the change in pricing models as organizations move from consumption to per seat? And do you even feel that given the way your business is put together?
Honestly, I think there's much more debate in the investor base than I see any tangible impact right now because ultimately, we don't price for cost plus or users or consumption. We price for value. And it's in our control to price for value. And it doesn't matter if you sell a 100-seat at $10 or 10 seats at, no, I have $100. So with that, I think we have a lot of flexibility. We have per seat and consumption meters. We have the technical capabilities to do both where adequate and to also change commercial models where adequate, where the market has changed or will change.
But I do also think that the subscription per user per month model, in a way, the sun will set a lot slower than something because of the reasons we just discussed earlier, the cost control, the budgeting, the very hard, boring facts of IT procurement. Consumption always sounds super attractive, but then the people who actually underwrite the bill, they really like a predictable budget. And we've seen actually that journey in AI too, where initially everything was consumptive. Suddenly, you saw some of the most successful, most pervasive copilots, they are not on a consumptive meter per se today. So it's really, I think the important thing for us is to have the flexibility and capability to go from one to the other or hybrid as the market evolves with the technology. And there I feel very, I mean, ultimately, we sell technology to make that possible.
I feel very confident about our ability to, A, have the right meter for the right, so the right hammer for the right nail, also in the long run, and second, ultimately, to price for value. That will protect us from. It doesn't matter then if it goes from seed to consumption and back or hybrid, as long as we are able. There we've done so much, and I'm so proud of what we've done on the go-to-market transformation to move from, okay, we sell you a contract and are gone. I'm exaggerating now to move to really a value selling and continuous customer success approach with customers. I feel we are in a very good spot when it comes to the monetization of AI in the future.
Sebastian, in the Agentic AI era, we talk about agents talking to other agents. Then, with the role of the orchestration layer where SAP has, you are at the core of your customer system. Can you please share with us your view on that orchestration layer and the role that SAP will play in that context?
Yeah. I mean, first of all, within SAP, of course, you need to orchestrate the multiple agents across our suite. And I think that's just another example of these types of platform investments that will give us, in my mind, a pretty unique edge against some of our peers because all of that takes R&D. It's not that this is a simple fix. I mean, as the industry aligns on A2A and the protocols and how to do this. But we feel confident. We feel confident that our concept of what I would call dual assistance is a great path forward. For that, you need the orchestration. You need multiple agents to orchestrate around a user, be it in controlling treasury or wherever else. Then, of course, the next level is to orchestrate those assistants to talk to each other where we are also having some very interesting research.
And then also, of course, how to exchange, how to do agent-to-agent communication across SAP and non-SAP solutions. And I think there in the past, I mean, we have been always open to integrate with the right technology, and that has served us well and will continue to serve us well. But I feel this is an area of intense research where we are embedded into all the key circles of where that research is happening. I think we do all the right investments there in our central AI team and then diffuse that into our different areas of our suite. So that's a very good concept that we've seen a lot of scale and success with so far.
Perfect. I'm going to follow up on something you said a little earlier. Where does BTP and Business Technology Platform and Business Data Cloud, for those who don't know it, fit in within the AI-centric world? And do you see SAP redeveloping over time the core functionality using AI, or do you see AI more as a way of extension of the functionality?
We love our acronyms. So let's start with BTP, actually. I mean, BTP is our Business Technology Platform. It's the common foundation for our own applications, but also for extensibility and integration of our estate with either custom solutions or third-party solutions. We've done massive progress over the last five years. I mean, many of you will know we came from a legacy of many, many acquired companies all running on their separate islands. And we were a best of breed under almost under one company umbrella that we have flipped now completely to we've actually done a lot of the heavy lifting work so that customers can experience a best of breed as a suite with SAP. And that, of course, now helps significantly with AI because all our foundational AI capabilities live on that same platform. Think of it as a path that's opinionated to SAP.
That's where we build our agents. That's where we actually build the agents that we deploy for our standard solutions. That's then also the toolkit that will be available for customers to extend our agents and to build their own agents in an SAP-specific context. This will be, I think, the work we have done there over the last five years will actually pay off massively for AI because now, I mean, most of our applications sit on a common database. We have done a lot of the underlying work to integrate, to harmonize workflows, UIs. All of that will make it easier for us to deploy AI into this stack because otherwise, you have to find six different ways of how to deploy agents into the software stack that you are running or 10 or 20.
So that's BTP, I believe, a huge asset for us when you think of it from an AI way, but also when you think of customers wanting to move to a Clean Core, removing modifications from their estate. So significantly growing faster than SAP average and I think set up for very nice sustainable growth. Then BDC is our Business Data Cloud. We launched that in February. This is our answer. I mean, I believe the next wave of AI, it's not only about adoptions. It's about how to bring business data into AI. And this is our bespoke answer of actually how to do that in an SAP-opinionated way, how we are now building data products into every of our applications, how to expose them into a data platform, and with that, also make them available as tools for the agents.
And through the partnerships we've struck with Databricks, Snowflake, Google, Microsoft, we can now actually extend that to non-SAP data, which we couldn't do in the past. So that would make it even richer. And it's actually pulling a lot of data gravity back to SAP. So those two, for me, it's the holy trinity. With every customer that I'm discussing, a strategic partnership, it's not just a RISE with SAP move to cloud. It's always also AI, BTP, and BDC as part of the package because that's then really where the full value of this apps data and AI flywheel lights up.
Excellent. I want to ask one of the questions that I get constantly, the bear case on software today, enterprise software. And that is, what do you think of the theory out there that IT organizations are going to use AI to build their own enterprise app or even extend enterprise apps? Because once you expose, in theory, your data externally, what would be the why wouldn't it happen? And what do you do to make sure that you capture that value?
Yeah. Interestingly enough, I see the opposite from happening right now, at least in many of the domains we are active in because what IT departments also learned over time that the lure of put a workflow on top, put an RPA on top, put an agent on top, custom build something, oh, it's super quick. Let's just do it. It's quick in the POC stage. It's never quick in the deployment, continuous optimization, training. So I think as long as we do a great job of shipping embedded AI, embedded agents with our full stack, and I see more and more customers that much rather put the pressure on us than using their precious IT resources. Not many IT departments can afford to deploy hundreds of people who can vibe code the next multi-agent system.
So they want to focus those precious resources in the area that matter the most to the company. And actually now, the second part is, so actually companies are quite happy when I talk to customers to tell us, "Hey, show me your finance roadmap. Show me your HR roadmap. Show me your procurement roadmap." It's actually the reason why they buy our software in many cases at this point, the agentic roadmap. But then they also do understand the right hammer for the right nail. Just because you can do something with agents or workflow or whatever it is, it doesn't mean that you should do it. I mean, you look at some of the most modern AI-forward startups, they still somewhere have a relational database because some transactional tasks don't lend themselves well to. It's just not the right technology to solve that problem.
And then the interesting piece is now I do see a huge interest as long as we deliver a strong agentic roadmap to use the full stack from SAP in those areas. But I do also see an emerging interest to work with us then in those other areas, those high-value areas like in supply chain, manufacturing, to bring our vertical knowledge together with the platform capabilities and the partnerships we've struck in areas that might in the last five to 10 years. That's Bernstein's copilot calling you.
No worries. To work with us in these bespoke high-value use cases, an area where I would say in the last couple of years, you wouldn't have looked to SAP for to work with when it comes to very complex bespoke use cases, but where we have a lot of heritage, a lot of IP that we can bring to the AI party that customers, I think, have more and more interest to work with us, so.
So to summarize a bit, the belief is that because of the unique expertise and the unique capabilities and the center of gravity of data and that contextual information, the AI is far more likely to be done internally versus externally. Is that a fair way to think of it so long as you continue to deliver quality AI capabilities your own? If so, what then drives the clients to bring more external data into SAP and run what could be run externally in terms of is it the fact that they're bringing that data in order to use it in conjunction with SAP data?
Exactly. I mean, if you look at many high-value use cases, take supply chain, that customers have built in the past prior outside of SAP's four walls, that's probably the wrong way. I learned that that doesn't work in English. Often these use cases have 50% or more SAP data in them. What happens then is, in the past, what happened is customers had to copy-paste the entire SAP data into a data lake. And then they had to replicate all of that data semantics somewhere else. And sometimes that worked. And oftentimes maybe that also didn't work because, I mean, if that would be so easy, SAP would have been gone 30 years ago. I mean, that's the core of our IP, how the data semantics, the process context, I mean, the source code of our applications.
And of course, once the data gravity sits with us and it's on our platform, we can much more control how to make some of these unique assets available to the platform without the complete data leakage to somewhere else, but making available in an opinionated way then also some of these partner capabilities, for example, to make third-party data accessible to SAP. So I think there's this misconception, oh, Business Data Cloud might now become a data donor to someone. I think it's the opposite. I mean, we have been a data donor to the world for a long time. And by the way, the data is always owned by the customer, not by us.
But actually, it's more of a recapturing data gravity by not doing something unnatural, but by providing unique value on a fully integrated stack that will never be replicable in a third-party solution, so.
So the expectation is in most instances, while you may ship some data to Office 365 Copilot to answer a specific question, by and large, the natural direction is the data will flow into SAP, you believe, from the other systems?
I'm not sure if that means that data flows in from other systems to SAP.
Yes.
But I think that based on the platform progress we've made both on the data side as well as on the technology side, that it's become far more attractive to use our vertical knowledge on an SAP-centric stack for use cases that are dominantly leveraging SAP data. That doesn't mean that we are now going out to compete in areas where there's no SAP data gravity in. But in cases where you are using SAP data predominantly, take supply chain as an example, data that sits in our system, I mean, what would you prefer? Replicating the data logic somewhere else or doing that natively and with a lot less effort in SAP? So the answer is clear to many customers. Again, then it's up to us to deliver enough data products, to deliver enough agentic extensibility capabilities.
But I think that's something I feel very confident in our capability to do.
Sebastian, I want to touch on another topic, completely different from AI. When we think about SAP, we think large organizations. But you have a huge market with small and mid-market enterprises out there. Could you please share with us what your plans there, what you're doing, and what's the prospects there for you?
Thank you. That's a huge passion of mine. I think while I get many investor questions about, oh, how long can that migration journey last? Much longer than you think. Actually, what's an underappreciated part of the story is the significant progress we've done on winning net new customers in the mid-market and winning actually some of the hottest, fastest growing AI companies in the industry in the world to bank on SAP as their bedrock foundation to scale to IPO and infinite scale, whatever their plans are. And that's something that makes me super proud. I mean, take just recently Snowflake, Databricks, Perplexity, Mistral, and on and on and on. And I think what has happened is we have matured our cloud ERP product in a way that it can be implemented in weeks, not months.
But at the same time, it gives you the scalability of something that can take you to become one of the world's most valuable companies. You never need to switch again. And that's a very, very attractive value proposition that works actually globally. It's not just a U.S. or European thing. It's actually we see very healthy growth in Asia and all parts of the world there. And I think that's something that will only accelerate as our ecosystem gears up to support this journey more and more because a lot of that will be partner-driven. So it will not be SAP salespeople now going into every corner of Silicon Valley, into every corner of maybe India, into every corner of. It's actually partners that will do that.
We've done a lot of work how to unlock that from incentive structures, from marketing approaches, from partner just tech infrastructure on how to enable them to take our solutions to market. We see very, very healthy growth in that sector. Then I think when you think about the opportunity there, I think 60% roughly of our TAM is in the mid-market, but only 30% of our top line is there today. There is a very, very good case to participate much more in that market moving forward.
You made the statement on partners. Can you give a more color on how you think about where partners fit in in terms of bringing the deals in and for these SMBs, how far they go in terms of getting the product operational and everything else?
Yeah. I mean, we have many different types of partners. Now, specific to this SMB mid-market segment, of course, one is resale partners that actually take over not only the selling itself. They take over the presales. They take over the contracting. And that took some work for us, but that's done now to actually enable them to do all of that autonomously. And then many of them then go on to do the implementation itself. And that's a very attractive business model as this part of our business will continue to grow significantly. And I see partners investing to up capacity across the entire value chain to support this motion. But then, of course, we have many other partners, right? The SIs. We have software partners. So we have a huge and quickly growing ecosystem that is supporting SAP.
Beautiful. I want to make sure I hit this question from the client. What do you from the audience, I should say, what's the biggest risk to SAP over the next decade?
Decade? Wow. I sometimes wonder if I can predict the next week. But now decade will make me think harder. I think, look, I want to go back there to the what and the how. I feel very confident in our ability to compete at market pace or ahead on delivering the what, the right solutions with agentic AI natively embedded with a strong data platform. Of course, there's a lot we need to deliver, but that makes it exciting. I'm not afraid of something like AGI eating SAP away tomorrow. I think that's a fallacy. Now, the how is the interesting part. I mean, the key reason that's holding back customers today from adopting our AI in the cloud is the installed base that we have.
So for me, a huge focus and a big area of investment for us is how to actually accelerate the migration journey, how to use AI and our tools like Signavio, LeanIX, all with embedded AI to help accelerate the migration to cloud. Because, of course, if you look at it, a customer who's not in the cloud, who doesn't have access to our agents, who doesn't have access to the latest innovation from SAP, of course, then at some point, you look at other alternatives. So I think as long but then with the customers where we have modernized with them in a Clean Core cloud estate, I do see they are absolutely putting SAP first in AI adoption in the functions they are working with SAP.
So that's for me a huge area of focus, but also an area where AI itself can provide a lot of I mean, at Sapphire, we talked about being able to accelerate migrations by 35% in the time and cost. And I think that will only get better from here. And that will limit the hurdle to do the right thing and actually modernize your full stack and not just put something on top in the hopes that it will be good enough.
You don't see that because I think it needs to be asked, even though I know the answer to this. You don't see the AI startups. You don't see the LLM vendors being any type of threat to the core business. It's a matter of how fast you can run and how well you can run.
So first of all, I have to say I love a good competition. I mean, of course, only in an ethical and fair way. But that's what makes this industry so great. It's the most competitive on the planet. And I think we have shown over the last five years that we can run much, much faster by now. So I feel good about the assets we bring to the race in terms of context to AI, process context, data context. I feel good about the platform capabilities we by now have developed to actually run the race. Now, of course, there's competition. I mean, in COVID, there was a boom of SaaS 1.0 players. Now there is a boom of SaaS 2.0/AI players or SaaS 3.0/AI. I stopped the counting. I mean, we have been in this highly competitive industry not since yesterday.
But I think as a team, as a company, we have come a long way in our agility and ability to deliver to a quickly changing environment. And that gives me a lot of confidence as I look into the future that this will ultimately turn out to be an opportunity much more than a risk to SAP as the next wave of AI will focus much more on the adoption than on the technology itself. And it will turn out that this actually takes a lot more than just a very well-trained LLM.