Navigating the contemporary enterprise data landscape can sometimes feel like piloting through the legendary Death Star trench. From containerized data products to file-and-table formats and from zero-copy to layered persistence; the days where the most exciting thing you would find in an SAP BI landscape was an extensive Composite Provider are behind us. Over the past few years, we have witnessed an incremental paradigm shift in how organizations approach their IT and architecture strategies. SAP’s Data and Analytics portfolio, spearheaded by solutions like SAP Business Data Cloud (BDC), SAP Analytics Cloud (SAC) and the HANA database (along with the Knowledge Graph and AI Core), has been at the very forefront of this evolution. Let’s have a look at how SAP’s AI-Native North Star Architecture makes this shift tangible.
The paradigm shift
Traditionally, the focus for data and analytics was on consolidating data streams, optimizing storage formats, and ensuring robust reporting capabilities for executive decision-making. SAP operated under the banner of the Intelligent Enterprise, which focused heavily on the automation of highly repetitive tasks and basic predictive analytics. However, in the new reality of the Autonomous Enterprise (as presented by SAP during Sapphire 2026), merely storing data and generating retrospective reports is no longer sufficient. Businesses today demand systems that do not just execute predefined, hardcoded tasks, but are inherently capable of reasoning, learning, and adapting to new variables in real-time.
To bridge this gap, SAP introduced its AI-Native North Star Architecture. This blueprint is not just a high-level product roadmap; it is a fundamental reimagining of SAP’s entire enterprise software foundation. In the context of Walldorf’s Data and Analytics portfolio, the North Star Architecture is so crucial because it sets the strategic direction for an ecosystem where AI agents, complex business processes, contextual intelligence, and strict semantic governance converge. Without a unified architectural vision, integrating advanced workflows and AI into deeply entrenched, highly complex legacy ERP environments would inevitably result in fragmented data silos and risky AI hallucinations. Let’s unpack all of that.
A new architecture approach
At its core, the North Star Architecture brings multiple distinct layers together: a centralized platform layer, a robust data foundation, an intelligent process layer, and a unified user experience driven by autonomous agents (e.g. Joule). The North Star Architecture serves as the guiding framework that aligns every component of SAP’s solution portfolio, ensuring that enterprise data is no longer treated as a static asset, but rather as a fuel that powers an ever-changing organizational (IT) infrastructure. By systematically decoupling storage and compute while enforcing strong semantic governance, this architecture prepares SAP solutions for a future where generative AI and intelligent agents are embedded natively into the (business) fabric of everyday operations.
SAP’s primary goal here is to create an overarching semantic map of a business through the SAP Knowledge Graph. This graph encodes nearly fifty years of SAP ERP engineering, business semantics, and custom code (Z…) into machine-readable relationships, acting as the critical bridge that allows AI models to understand the intricacies of enterprise data without losing business context. Having said that, building an AI-native ecosystem requires more than just clever tools; IT landscapes will have to overcome the ‘data tax’ associated with constantly moving, replicating, and transforming data. I see this reflected in ever-growing discussions at customers on where (SAP) data is stored, where it is processed and where it is to be consumed. Enter SAP’s 2026 string of strategic acquisitions.
Understanding SAP’s acquisitions
In the above image illustrating SAP’s North Star Architecture layout, I have plotted the position of their three acquisitions in the Foundation Layer; right in BDC’s ballpark. But what do they actually do?
To ensure that the data foundation underlying the North Star Architecture is properly structured, SAP acquired Reltio earlier this year (in May). Reltio’s advanced master data management (MDM) capabilities bring Model Context Protocol (MCP) support for event-driven data sharing directly into the SAP Business Data Cloud, essentially creating a powerful AI-supported metadata catalog (sounds familiar…?). This ensures that the foundational entities AI relies upon are clean, harmonized, and form a single source of truth across the landscape.
Following this, SAP aimed to address the persistent headache of data silos by acquiring Dremio (in June), a move that effectively tackles the overhead of ETL processes. Dremio transforms BDC into an Apache Iceberg-native enterprise lakehouse and, coupled with the open data catalog Apache Polaris, this acquisition allows SAP and non-SAP data to be queried seamlessly via open standards without costly and historically required data movement. People who have already worked with BDC’s Delta Share functionality will notice that this is a step towards more agnostic data integration: where Delta Share is tailor-made for Databricks’ Unity Catalog and Spark integration, Iceberg allows customers to query and federate data via REST catalogs across multiple engines, such as Flink or Trino. BDC will thus offer both Delta Lake and Apache Iceberg table formats going into the future, allowing customers to leave their data in BDC and federate it to wherever it needs to be (either via Delta Share or Iceberg).
While Reltio and Dremio solve the distinct challenges of data governance and physical location, SAP’s €1 billion commitment to Prior Labs (whose acquisition was completed last month), specifically addresses the Model Services in the Foundation Layer. Standard Large Language Models (LLMs) often struggle with the mathematical precision of relational records; but businesses primarily run on highly structured tables, not just text. Prior Labs offers the solution in that it specializes in Tabular Foundation Models (TFMs) (e.g. TabPFN-2.6). These statistical reasoning engines learn directly from tabular data, offering lightning-fast zero-shot predictions and in-context learning. By embedding TFMs alongside its own Relational Pretrained Transformer (RPT-1.5) technology into the North Star Architecture, SAP ensures that its autonomous agents natively comprehend the statistical causality within structured business metrics. Hence, put together, Reltio, Dremio, and Prior Labs enable an important part of the North Star Architecture when it comes to data: harmonizing it, federating its access, and mathematically reasoning over its content.
AI infused across the board
The integration of the North Star Architecture and these recent acquisitions perfectly illustrate SAP’s broader AI philosophy: moving beyond simple read-only copilots toward read-write autonomous agents (Business AI, Joule). We can also see this reflected in SAP’s Data and Analytics offerings. For instance, in SAC, we are seeing the introduction of autonomous Joule Agents embedded directly into the SAP Enterprise Planning functionalities. These agents are capable of autonomously detecting external market data signals, simulating their quantitative impact on KPIs, recommending strategic actions, and orchestrating downstream planning updates. If we then take a peek into the future for BDC, the combination of the Knowledge Graph, AI Core integration, and Prior Labs’ TFMs will enable these agents to generate complex ‘what-if’ scenarios based on contextualized, real-time data. Furthermore, SAP is also leveraging Small Language Models (SLMs) to handle specific, highly optimized tasks within the BAIP, ensuring faster inference and significantly lower compute costs without sacrificing accuracy (read about these and the difference with RAG here). It should be noted that, today, most of this is still theoretical or in the development phase. However, I do expect SAP to start rolling out these functionalities in a shorter timespan than we’re used to, given their big focus on AI enablement and the recent finalization of all three aforementioned acquisitions.
The pursuit of EU-based digital sovereignty
Yet, as AI adoption accelerates rapidly, a critical challenge remains for many organizations, particularly those operating in the public sector and highly regulated industries: digital sovereignty. SAP recognizes that enterprise AI cannot be a black box hosted entirely on foreign servers (e.g. US), which is why they have made a visible push toward EU-based data and AI sovereignty. To realize this, Walldorf successfully cleared major regulatory hurdles, supported in large part by its recent €3.5 billion bond issuance designed specifically to fund a triple AI investment strategy.
Another important cornerstone of this sovereign approach is the strategic partnership with the privately held and France-based Mistral AI. Delivered as an out-of-the-box solution labeled ‘Mistral on SAP Business AI Platform’, this joint offering guarantees that AI workloads, including data, prompts, and model execution, remain strictly in-region. It operates on a consumption model with a unified AI Unit metric and a single SAP contract, which reduces legal and operational complexity for customers. Mistral on SAP is explicitly designed to comply with stringent frameworks like the EU AI Act, GDPR, and NIS2, in order to inspire confidence with (potential) customers. While SAP has also integrated Anthropic’s Claude to provide a diverse, best-in-class model ecosystem globally, the Mistral alliance is a decisive statement that I believe will resonate with an increasing number of non-US-based customers. If anything, this move demonstrates SAP’s support to an ‘AI made in Europe’ paradigm, empowering businesses to innovate with the latest and greatest in technology, while retaining control over their most critical company (data) assets.
Bringing it all together: the reality of the autonomous (AI) transition
If we look at the complete picture painted by SAP over the course of 2026, the underlying narrative is one of systemic transformation. SAP is reconstructing its entire ecosystem around the AI-native North Star Architecture; Sapphire’s message was very explicit in this. By acquiring Reltio to tightly govern master data, Dremio to federate an open-standard lakehouse in BDC, and Prior Labs to bring specialized Tabular Foundation Models directly to relational data, SAP is building a holistic engine that will power the continued emergence of agentic AI. Coupled with the strategic deployment of Mistral AI for robust European digital sovereignty, the high-level vision for the Autonomous Enterprise is shaping up to be technologically robust and (for now) flexible where many customers will wish it to be.
However, we must ground this lofty vision in the practical, hands-on realities of the contemporary enterprise data landscape. The North Star Architecture is a clear destination, but the journey towards realizing something like it (in-or-outside of SAP) will be a long one for most companies. A significant portion of SAP customers are still grappling with fundamental modernization challenges that preclude immediate AI adoption. Many organizations are currently heavily invested in mapping out complex migration paths, often leveraging temporary, intermediate stepping stones like BW 7.5 PCE, before they can practically consider adopting the more advanced features that BDC offers. While autonomous agents and zero-shot predictive models sound incredible on paper, they are completely dependent on the quality, structure, and accessibility of the underlying (master)data; which has been the age-old challenge that has validated the existence of consultants like myself for years and remains a fundamental condition for sustainably competitive AI use.
Thus, for the North Star Architecture to truly deliver on its promises, at the very least in the domain of data and analytics, organizations must first meticulously clean house. The Autonomous Enterprise will not emerge overnight through a simple software update or a flip of a switch; it will require rigorous data engineering, a strategic organizational commitment to transitioning away from legacy (BW) structures, and a fundamental review of how organizations integrate data pipelines and maintain their business semantics and context. Aside from this, we can cannot forget the cost-aspect of it all: although some of SAP’s agentic features are free today, Joule Premium and updates to the per-user-per-month pricing model are quickly changing this. Proof of value thus is crucial, and I will remain highly skeptical of AI-based outcomes until they have been rigorously tested, not to mention the less visible cost considerations (e.g. natural resources) I think more people and companies should make. All in all, I think that SAP has laid quite some solid groundwork for the AI era we are all (already) living in today; only time will tell how soon and how thoroughly all this technology can deliver on all the promises.
If you want tailored advice and quality in-depth support to continue your organization’s data and analytics journey, do not hesitate to contact Expertum. Alternatively, feel free to read our other blogposts on SAP Business Data Cloud for more insights that might help you with just that little problem your implementation is facing.