Clear data architecture makes AI effective

E3 Magazine: Mr. Hermanns, AI is at the heart of the new SAP world. How important is AI for businesses today?

Lars Hermanns: Artificial intelligence has now become a strategic issue for businesses. The pressure to engage with AI has increased significantly—partly due to the activities of major software providers like SAP, which are increasingly expanding their solutions to include AI capabilities. At the same time, we are seeing a high willingness to invest among our customers. AI is viewed as an essential component of future business models.

E3: How far along are companies in the practical implementation of this?

Hermanns: Many companies are already using AI in specific areas, such as to support analytics or automate individual processes. However, comprehensive integration is rare. We often see pilot projects or isolated use cases, but there is a lack of a company-wide strategy and a scalable technical foundation. Furthermore, there is an expectation that AI can do everything.

E3: What is preventing widespread implementation?

Hermanns: In many companies, data has accumulated over time, is scattered across various systems, and is often inadequately documented. Pilot projects frequently rely on cleaned and perfectly formatted data—conditions that are rarely replicated in real-world operations. The productive use of AI often lacks clear structures, consistent definitions, and transparency regarding which data is trustworthy. Without this foundation, it becomes difficult to reliably scale AI applications.

E3: What are the basic requirements?

Hermanns: There are two key areas of focus. First: data architecture. Many companies operate with fragmented system landscapes in which data is redundant and, in some cases, contradictory. The goal should be to establish clearly defined data products that are structured, quality-assured, and described in technical terms. Modern architectures such as lakehouse approaches with multi-tiered data models—for example, Bronze, Silver, and Gold—can help here. Second: data governance. Companies need clear responsibilities for data—both technical and business-related. Roles such as data owner or data steward are crucial for ensuring quality, consistency, and context. Without governance, there is a risk that AI will access incorrect or misunderstood data and thus deliver flawed results.

„Successful companies integrate AI where,
”where it solves specific problems.”

Lars Hermanns,
Head of SAP Analytics,
Nagarro

E3: What role does the data fabric play here?

Hermanns: A data fabric can be understood as an overarching architectural principle that integrates data across different systems and makes it usable. At its core, the goal is not only to physically consolidate data, but also to intelligently link it through metadata, semantic models, and unified access layers. This makes it possible to consistently access data regardless of its storage location and make it available for various use cases—including AI. A data fabric complements existing platform concepts and provides a unified view of distributed data.

E3: How can data be improved for use with AI?

Hermanns: It is crucial that data is not only technically accessible but also conceptually understandable. Data is organized into structured layers and undergoes quality assurance. In the „Gold Layer,“ this results in curated, domain-modeled data products. These are often based on established modeling approaches such as star or snowflake schemas. In addition, metadata and catalogs are created that describe the origin, meaning, and use of the data. This context is particularly crucial for modern AI applications. Data can only be used effectively if it is clear what it means and how it is related.

E3: What about the human factor?

Hermanns: Companies should leverage the valuable knowledge their employees have built up over the years. When employees„ intelligence is combined with company-wide process knowledge and adaptive AI models, it creates a powerful catalyst for transformation that goes beyond technology alone. At Nagarro, we call this “Fluidic Intelligence.”.

E3: How can AI be integrated into the strategy and existing structures?

Hermanns: In a Fluidic Enterprise, AI is viewed as an integral part of existing processes and decision-making structures. Successful companies integrate AI where it solves specific problems—such as in decision-making processes or when analyzing complex relationships. The focus is less on disruptive changes and more on the targeted refinement of existing structures based on a solid foundation of data and architecture.

Continue to the partner entry: