In this comprehensive interview at SAP Connect in Las Vegas, Stephan de Barse, President of Global Business Suite at SAP, makes the case for why best-of-breed enterprise software is being replaced by best-of-suite approaches, driven by the commoditization of the application layer through AI.
De Barse, who previously spent seven years at o9 Solutions, a best-of-breed supply chain vendor, brings a unique perspective to the debate. The discussion covers SAP’s AI strategy centered on Joule, the challenges of multi-vendor environments, agent-to-agent communication standards, and how organizations can manage the proliferation of AI agents.
This transcript has been organized by topic to help readers navigate the conversation and find insights relevant to their interests in enterprise AI, application consolidation, and digital transformation strategy.
Also read: SAP opens platform with MCP: AI agents can communicate with SAP
The end of best-of-breed announcement
Question: SAP sent out a press release and said during the keynote it’s the end of best-of-breed and it’s now all best-of-suite. You stand by that messaging?
Stephan de Barse: I do, wholeheartedly. And I came from a best-of-breed solution provider. So I spent seven years as a best-of-breed solution provider in supply chain. I have seen best of breed, now I’m best of suite. So I feel I can contrast and compare in a good way. But I absolutely stand by that.
Question: Which solution was that, just for the record?
Stephan de Barse: It was o9 Solutions, and it is a supply chain best-of-breed vendor in the supply chain space.
Question: Why do you believe that best-of-suite is going to be the next thing, and it’s the end of best-of-breed?
Stephan de Barse: I think, look, our point of view is that very similar to what we saw in the infrastructure layer that got commoditized by the cloud, obviously still critical, but commoditized. We believe the app layer will be commoditized as well, and most of the value will sit in what you can unlock with your data. And then how is AI going to help you in terms of managing your business better and making better decisions? Because if you think about it, best of breed solution providers, they came up oftentimes because they offered a slightly better user experience, or there was a feature that they liked better than another feature. Now, what then happened is you now have a lot of big organizations that have a very desperate, proliferated enterprise landscape.
AI and the appless experience
Question: How does AI change the competitive landscape for best-of-breed vendors?
Stephan de Barse: In the world of AI, the entire user experience is not so much about the app layer anymore, it is about what we call Joule. So it was interesting actually, yesterday ChatGPT announced that you can now access consumer apps, booking.com, Figma, and Retro Alliance. But with Joule, we’re going to see exactly the same. And that’s why yesterday, when we did the keynote, Mohamed talked about the appless experience. So as a user you don’t have to go through all those different apps. The entire experience will be through Joule. So I think that’s one of the reasons that best-of-breed, I think lost a lot of their differentiation because they will not be picked based on a better user experience or feature here or there because everything will be commoditized now by AI.
Stephan de Barse: The Second big one is if you think about AI and delivering AI at scale, if you’re an organization and you run your business on thousands of different applications and all those different applications are trying to do something with AI, that becomes just an absolute integration nightmare.
Data as the foundation for AI
Question: If you have a best-of-breed solution with the most features and actions, AI can leverage those features and actions. If you take them away, it cannot necessarily leverage them. Who’s building the AI then?
Stephan de Barse: At the end of the day, we are a solution provider, so we build apps. Those apps do two things. They run an end-to-end business process, and they generate a lot of reliable, trustworthy data. Now, then, the key question is what do you do with the data? So there’s SAP data, which we provide with two big motions. There is no copy, so I don’t have to copy the data. And then rich semantics, what do we mean by that? Take a sales order as an example. Sales order has customer data, it has material data, it has delivery data, it has all that data. Now, if you had to recreate that, you would then need to pull data from 15 different SAP tables and recreate that logic. So that’s just very cumbersome to do.
Stephan de Barse: So now we say SAP data, you combine it with non-SAP data, and that is then the foundation for AI. But our AI strategy is all about embedded AI and agents that make it back to the core application. Five weeks ago, this article came out from MIT: 95% of Gen AI pilots failed. So one of the big reasons is that if you have some AI capability as a layer on top that is completely disconnected from the end-to-end business process, it will not scale, and ultimately it will not drive a change because ultimately it needs to be brought back to where the transaction and decision happen.
Multi-vendor reality and integration
Question: This vision is like an ideal world in theory, because in practice, it will never work because enterprise organizations always buy multiple vendors. For example, if an organization is now heavily invested in SAP and Salesforce, and maybe Workday, you’re aiming to have them ditch Salesforce and Workday and go all in on SAP. The chance of that happening isn’t really high. At the same time, you’re going to support MCP and A2A for agent-to-agent communication. But that’s about agents working together. So it’s a contradiction to say we’re going best-of-suite, but we all need to work together.
Stephan de Barse: I would say it’s not a contradiction, because we acknowledge that only very few companies will be 100% SAP. So we acknowledge that. At the same time, what we also acknowledge is that there’s companies that do not have three or four or five different applications. They have thousand plus different applications. That is just unworkable, especially when we think about AI. So there is going to be a big harmonization taking place in the app space. We are seeing that already.
More and more of our customers are migrating to many of the different SAP solutions in the app space because they understand that we then bring the data together. SAP and non-SAP data, we can deliver a unique AI experience. Now, to the point you’re making: yes, we do acknowledge that there will be a Salesforce, an Adobe, a Workday, and other solution providers. And then obviously it is important that agents can work with agents from other software providers. What we do not believe is that a company will consume thousands of agents from thousands of different solution providers. We don’t see that happening. It’s not practical, it’s extremely expensive. And the integration nightmare will be quite significant.
Managing AI agent proliferation
Question: There’s another issue I see at all the vendors at the moment regarding SaaS solutions and AI. We’re getting AI agents in every solution. I wonder how a business user is going to have an overview and prevent this AI chaos because you guys introduced, I think, two dozen AI agents this week. People can choose from 100, 200, or 300 agents, depending on their software portfolio. How are they going to figure out when there’s an agent, which one to choose?
Stephan de Barse: This is exactly the argument that I’m making in favor of the business suite as an organization. These agents and these capabilities need to work together. However, to that point I agree with you. It cannot be a hundred, two hundred, three hundred different applications all providing agents, and then you need to work together. That doesn’t work. The other thing that we have seen that does not work is if you take procurement to pay and say there are 15 steps in procurement to pay, you take one step and you say oh, there’s this agent I might use for that. And this other step, there’s that—this becomes a complete chaos. That’s why we are saying we need to bring it back to the core business process.
Question: But how is a business user going to find these agents? Because there are so many.
Stephan de Barse: As a business user, it comes back. So you would use an application for, let’s say, procurement. Then within procurement, we would then have embedded AI cases and agents for the user to work with. What is not happening is that you take procurement and say, oh, by the way, I can now use 20 different agents from 20 different solutions. That’s not happening. We go to our clients and say, okay, you have been using Ariba in a certain way. We announced next generation Ariba. Then on top of that we have AI cases and agents and we work with our customers to reimagine those processes. And then it comes back to how do you then support them? And that’s where the tool chain comes in. LeanIX will be extremely important for modeling and mapping future architecture, including agents and integrations. Signavio is going to be extremely important, which is not anymore just the process that we knew. But what is a business process that agents support? And then there is WalkMe that will also help.
SAP’s competitive advantage
Question: SAP wants to become the UI for AI basically for organizations and to control that. All your competitors want to do the same. What makes SAP win this battle? What do you have that they don’t?
Stephan de Barse: At the end of the day there’s two things that we have that none of our competitors have. We have the broadest portfolio of end-to-end business applications. At the end of the day, it comes back to where you run your end-to-end business process? And the second thing is, there is a lot of extremely valuable data that is needed to do any meaningful AI that’s sitting in an SAP system. And yes, you can take that out. You then need to recreate the semantics. The moment there is a change in your business, you need to recreate the semantics that you already created. It can be done. This is very time-consuming and very costly. So why would you not do that with the SAP portfolio?