After AI comes BI. Why AI doesn't work without a foundation.

"As a BI professional, I always say: after AI comes BI."

Usually, there is laughter, after which the conversation quickly continues about AI initiatives. Understandable, because AI provides direct answers and thus gives the impression of insight. It promises speed and less dependence on reports.

In practice, however, a different picture often emerges once the first applications go live. The AI response deviates from the dashboard. Definitions are incorrect. And when you ask how the answer was derived, there is silence.

What at first glance seems like an AI problem turns out to be something more fundamental. This is a data foundation problem.

AI makes existing problems more visible

AI is not a new data source. It is a new way to query existing data. Where you used to open a dashboard, you now ask a question to a copilot or agent. The interaction changes, but the underlying data remains the same.

And that's where the risk lies. AI presents answers as if they are correct. Convincingly formulated, without reservation. But if the data is incorrect, or if definitions are not consistent, then AI does not provide better insight. It delivers a credible interpretation of bad data.

AI does not accelerate insight if the foundation is not correct. It accelerates inconsistency.

What it looks like in practice

Many organizations work with multiple platforms alongside each other. SAP for business processes, Microsoft for collaboration and analysis, or a combination that has developed over the years.

That combination is powerful. But in practice, those worlds are too often separate. Data is copied between systems. Definitions are reinvented per department or tool. "Revenue" means something different here than it does there. And "customer"? That depends on whom you ask.

This is not an exceptional situation; it is rather the norm. And as long as that foundation is not right, no AI initiative will help. The technology is there. The foundation is missing.

The foundation: one semantic layer for everything

The solution does not lie in more tools. You already have those. The solution lies in one layer where definitions are central: what is a customer, what is revenue, what is a region. Once established, valid for reports, copilots, and AI agents alike.

This is what BI has been doing for years, but it is more important than ever in the AI era. Because AI does not work on raw data. It works through the meaning you give to that data. Without that semantic layer, AI is just a fast track to the wrong answer.

Which platform suits you?

The good news: you don't have to build this foundation from scratch. Modern data platforms are designed for this. And as an SAP or Microsoft customer, you have multiple routes.

• SAP Business Data Cloud brings SAP data together in one managed environment, with built-in semantics based on your ERP processes. Strong if SAP is the heart of your organization.

• Microsoft Fabric combines lakehouses, data warehouses, and AI applications within one integrated ecosystem. Strong if you are already deeply embedded in the Microsoft world and want broad integration.

• Azure Databricks offers maximum flexibility in data engineering and machine learning. Strong if you need customization and have a strong technical team in-house.

And sometimes the answer is: both. SAP and Microsoft are not mutually exclusive. On the contrary. The right choice depends on your existing landscape, your team, and your ambitions.

Back to the joke

After AI comes BI. Not as the next step, but as a prerequisite. The organizations that get the most out of AI are not those with the most advanced models. They are the organizations that have their data foundation in order. That know what their data means. And that have one version of the truth. For everyone, in every system.

Would you like to know which foundation suits your organization? We are happy to think along with you without a preference for a platform, but with knowledge of both worlds.