SAP AI North Star Architecture: A Guiding Vision for the Future or a Concrete Blueprint?

With its AI North Star Architecture, SAP outlines how enterprise software is expected to work in the future. But what does this vision mean for SAP customers? Is the architecture a concrete blueprint for their own AI landscape or rather a strategic direction? And what decisions should businesses be making today? Larissa Haas, Squad Lead “Analytics & Automation” at sovanta, takes a closer look at the architecture, its guiding principles, and the questions that remain open.

A North Star Is Not a Roadmap Yet

With its AI-native North Star Architecture, SAP has presented an ambitious vision for the future: enterprise software is set to evolve from a collection of individual applications into an intelligent, interconnected system. AI agents will no longer simply answer questions or support individual tasks. Instead, they will be able to understand entire business processes, prepare decisions, and act autonomously within defined boundaries.

This goes far beyond integrating generative AI into existing SAP applications. SAP describes a fundamental architectural shift: from a traditional System of Record to a System of Context, where data, processes, applications, and AI agents work together.

The architecture consists of four layers:

• User Experience Layer: Interaction with enterprise software becomes more dynamic, personalized, and increasingly driven by Joule. Instead of navigating between applications, users express their intent and receive the information and functionality they need.
• Process Layer: Business processes are executed through the interplay of traditional applications, deterministic workflows, and autonomous AI agents. Systems and agents outside the SAP ecosystem can also be integrated.
• Foundation Layer: Enterprise data, semantic knowledge, and AI capabilities form the intelligent foundation. Technologies such as SAP Business Data Cloud and SAP Knowledge Graph are designed to ensure that AI operates within the relevant business context rather than in isolation.
• Platform Layer: The technical foundation for developing, deploying, and operating AI solutions, including security, governance, monitoring, and lifecycle management.

These four layers do more than describe a new technology stack. Above all, they illustrate how SAP intends to shape collaboration between people, applications, data, and AI in the future.

But this is precisely where the key question for businesses arises: What can already be implemented today, and what remains a vision for now?

Blueprint or Vision: How Definitive Is the SAP AI North Star Architecture?

SAP itself describes the AI North Star Architecture as a blueprint for the Autonomous Enterprise. However, this does not mean that businesses can simply adopt the architectural components presented and automatically obtain a functioning AI landscape.

The AI North Star Architecture makes it clear where SAP intends to take its technologies: deeper integration of AI into business processes, more unified access to data and functionality, open communication standards for agents, and centralized governance and orchestration.

At the same time, the SAP white paper describes various capabilities whose technical implementation and availability are still evolving. This is where it quickly becomes apparent that important questions remain unanswered, particularly regarding how individual components will work together.

For example: How will custom-built agents be integrated into SAP orchestration in the future? What roles will Joule Studio, Agent Gateway, and existing automation tools play? How will permissions be managed across system boundaries? And what will the lifecycle of AI agents look like across development, testing, and production environments?

The answers to these questions are crucial for concrete architectural decisions, but not all of them have been conclusively clarified yet.

However, this does not mean that businesses should put their AI strategies on hold. Instead, it makes sense to distinguish between the long-term vision and decisions that can be made with confidence in the short term. The North Star sets the direction. Each business must chart its own path based on its existing system landscape, business processes, and requirements.

What Decisions Should Businesses Make Now?

In our view, the SAP AI North Star Architecture already provides a basis for identifying several fundamental questions that businesses should address today.

1. What Role Should AI Play in Our Business Processes?

Not every process requires an autonomous agent.

SAP explicitly distinguishes between deterministic and AI-native approaches in its architecture. Traditional applications and workflows remain useful where processes are clearly defined, predictable, and rule-based. AI agents complement these capabilities, particularly where information needs to be interpreted, different situations assessed, or decisions made flexibly.

Businesses should therefore not ask: Which process can we replace with an agent?

The better question is: Which parts of a process actually require intelligence, and which can still be reliably handled through deterministic approaches?

This distinction is particularly important in SAP landscapes where extensive automation is already in place. AI-native does not mean fundamentally replacing proven technologies.

2. How Can We Establish a Reliable Data and Knowledge Foundation?

An AI agent is only as good as the context on which it bases its decisions.

Here, SAP focuses in particular on SAP Business Data Cloud, data products, and SAP Knowledge Graph. The underlying idea is that AI should not merely have access to raw data, but also be able to understand its business meaning and relationships.

This raises the question of how businesses can prepare their existing data landscapes for this. Where is relevant data stored? How can it be made accessible across systems? Who is responsible for data quality and access permissions? Which information actually needs to be available to AI?

These are not new questions. However, the use of autonomous agents makes them considerably more important.

3. How Do We Integrate Custom AI Solutions into the SAP Ecosystem?

Few businesses start with a fully standardized AI landscape. In many cases, custom AI applications, external models, automation platforms, or specialized agent frameworks are already in place.

The North Star Architecture takes this reality into account. SAP relies on open standards such as the Model Context Protocol (MCP) for accessing tools and the Agent2Agent Protocol (A2A) for enabling collaboration between different agents.

This sends an important signal: the future AI landscape is not intended to consist exclusively of SAP-built agents. However, open standards alone do not guarantee seamless integration. Businesses still need to decide where to develop their own agents, which capabilities to expose as tools, and which platform should handle orchestration.

Distinguishing between existing workflow and automation solutions and new agentic capabilities will also be a key architectural decision.

4. How Much Autonomy Do We Want and Can We Allow?

Perhaps this is the most important question of all.

An agent that summarizes information requires different permissions and control mechanisms than one that independently places orders, modifies master data, or communicates with customers.

For such scenarios, the North Star Architecture provides for mechanisms including centralized access controls, agent identities, monitoring, and human-in-the-loop mechanisms. However, technology cannot make the fundamental governance decisions on behalf of businesses. Which decisions may an agent make autonomously? At what level of risk is human approval required? Who is accountable for its actions? And how can it be traced why a particular action was taken?

These rules should be established before the first agent starts making decisions in a production environment.

What Guidelines Does SAP Set, and Where Do Businesses Have Flexibility?

The North Star Architecture gives businesses considerable flexibility. SAP explicitly positions the architecture as open to partner technologies, third-party models, and customer-developed solutions.

At the same time, a clear architectural framework is emerging. SAP emphasizes standardized access methods, semantically enriched enterprise data, centralized governance, and increasingly unified orchestration. Joule is also expected to play a key role as a central entry point for interacting with applications and agents.

For SAP customers, this means that technological openness and architectural independence are not necessarily the same thing.

For example, businesses developing their own AI agents should already consider how these could later be integrated into an overarching orchestration and governance framework. Those relying on alternative data platforms should familiarize themselves with the intended integration standards. And businesses developing their own user interfaces should keep an eye on the shift toward dynamic, context-aware interactions.

However, these are not blanket prohibitions on alternative technologies. Rather, they are strategic guidelines that may influence future integration options and development effort.

Practical constraints also come into play, including existing SAP contracts, available services, licensing models, deployment options, and the individual system architecture. Determining which combination makes the most sense in the long term therefore requires more than the white paper alone can provide.

What We Recommend Businesses Do Today

Although SAP’s vision has not yet been fully realized in every technical detail, it already provides a basis for concrete recommendations on how businesses should proceed.

Conclusion: Our Perspective on the SAP AI North Star Architecture

The AI-native North Star Architecture is an important signal of SAP’s strategic direction. What we find particularly exciting is not the vision of individual intelligent agents, but rather the effort to systematically connect AI with business processes, data, and existing software architectures.

At the same time, the extensive discussion within the SAP community shows that several questions remain open between the architectural vision and productive implementation.

We will therefore continue to monitor closely how SAP further develops this vision. Which architectural components will actually become available? How will the integration of customer-developed agents evolve? And what opportunities will emerge for businesses that have already established their own AI solutions and automation landscapes?

For us at sovanta, one question is at the heart of it all: How can we translate SAP’s AI vision into a concrete, future-proof architecture that meets the needs of our customers?

Ultimately, the key challenge is not to replicate the SAP AI North Star Architecture as closely as possible. It is to use its strategic direction to make the right decisions for our own AI landscape.

And that is exactly what I would love to discuss with you personally. Feel free to get in touch.