SAP Business Data Cloud and the Business AI Platform after Sapphire 2026

We talk to many companies that run AI pilots on their SAP systems, and we keep seeing the same pattern. The pilot looks great in a demo. Then it never reaches production. The reason is almost always the same. The data underneath is not clean enough. It is not governed well enough. So, nobody trusts it for real decisions.

Sapphire 2026 was SAP's answer to this problem. Two acquisitions made that clear: Reltio and Dremio. Together they show where SAP thinks the real work needs to happen.

To understand the announcements, it helps to use SAP's own model. SAP calls it the AI Native North Star Architecture. It describes four layers:

User Experience layer. This is Joule. It is how people talk to the system.

Process layer. These are agents. They plan and do the work.

Foundation layer. This is the data and the AI underneath everything.

Platform layer. This is the runtime and the governance that holds it all together.

This blog explains what SAP announced at Sapphire 2026. We place each announcement in one of these four layers. At the end, we give you clear steps you can take now.

The four layers

So, let's dive into what the four layers are first.

User Experience layer

SAP wants Joule to be the front door for everything. The vision is that you do not click through apps anymore. You just tell Joule what you want instead. This does look strong in a demo, but this layer depends fully on the layers below it. Joule has been available in SAP Datasphere since early July but is still missing in other tools such as SAC (SAP Analytics Cloud).

Process layer

In this layer, apps become what SAP calls capability providers. Agents plan and carry out work across many apps. This is the boldest idea of the four. Think of an agent that handles a full procurement task by itself. No more clicking through five screens by hand. But this layer is also the least proven. Most agents shown at Sapphire are still previews. The AI Agent Hub, which should manage these agents, is not expected before the third quarter of 2026. And real use of Business AI in production is still very low at the time of writing.

Foundation layer

This is where the real investment went this year. It is the data and AI layer. And it is the direct answer to the pilot problem we described above. SAP said it clearly at the keynote. No AI agent can fix a bad data landscape. The North Star paper puts it this way: A system of record must become a system of context. That means it is not just about data anymore, but also the meaning behind it and how it connects. Only then can an agent reason about your business instead of guessing. SAP Business Data Cloud, or BDC, and its Knowledge Graph should make this possible.

Four new deals are present in this layer, which are as follows.

Dremio. This deal turns BDC into an open lakehouse, built on Apache Iceberg. So where your data lives, becomes less important. What matters is where the context lives, and that is BDC. This deal was closed on July 6. Dremio is now part of SAP. One thing to watch. Dremio is built on Iceberg and Polaris. SAP's Databricks partnership from 2025 is built on Delta and Unity Catalog. These are two different camps. If your BDC plan leans on Databricks, ask your SAP account manager how these two will work together in the long run.

Reltio. Reltio handles Master Data Governance inside BDC. Of the three deals, this one is furthest along. It closed on May 7. Reltio brings master data management across SAP and non-SAP systems. Duplicate customers. Suppliers spelled three different ways. No single version of the truth. These problems quietly ruin every report you build. Now they will also ruin every agent you build. This work is not exciting. But everything else in this layer depends on it.

Prior Labs. SAP is putting over one billion euros into this. The goal is structured business data. Things like churn, late payments, or demand forecasts. Not the text and chat tasks that normal AI models already handle well. This deal has not closed yet. SAP expects it in the second or third quarter of 2026. Even after it closes, Prior Labs will stay independent. It will not be fully absorbed into SAP. This is a promising idea. But it is not proven yet at the scale of a real SAP company.

Claude. Claude sits on top of all this as the reasoning engine. It lives inside the Foundation layer, but it does its work in the Process layer, through Joule. SAP called it a primary reasoning and agentic capability. Notice the word a. It is not the only one. Claude sits next to Mistral and Cohere. SAP calls those the sovereign options. It is worth pointing out something odd here. SAP talks a lot about data sovereignty. But one of its primary AI models comes from a US company. That does not fully match the sovereignty story.

So here is the honest summary. The logic behind this layer makes sense. The pieces are real. Two of the three deals, Reltio and Dremio, are already closed. Only Prior Labs is still open. This means the Foundation layer is not just a plan on a slide anymore. Most of it already belongs to SAP. The last piece should follow later this year.

What this means for you

Here is what you should do now, and what you should wait on.

Do this now

• Check how mature your data really is. This is the fastest way to see where you stand before you make important decisions that are hard to reverse. McCoy's Data and BI Assessment can do this for you.

• Run a BDC readiness check. Size the Capacity Units you would need. Model the cost against real, named use cases. Write down every assumption before you sign anything.

• Fix your (master) data first. This work pays off no matter what happens next. It belongs to the Foundation layer, and it is worth doing regardless of your BDC plans.

• Try one Intelligent App on your own data. Treat it as a small test. Nothing more.

Wait on this

Do not commit budget to the Process layer agent roadmap yet. The runtime is free until the end of 2026. That likely means a price jump in 2027. Wait until the AI Agent Hub and the agents themselves leave preview before you plan real spend here.

In short

Sapphire 2026 tells the clearest data and AI story SAP has told so far. The Foundation layer work is worth doing anyway, no matter what SAP does next. And the gap between the key-note promise and reality is shrinking. Two of the three Foundation deals are done, not just promised.

The main focus of most companies should not be on a new AI initiative but on building your data foundation first. AI initiatives success depends on it.