SAP's AI Strategy: Openness vs Control in Enterprise Agents

What is the core contradiction in SAP's new AI vision?

The enterprise software landscape has been undergoing significant turbulence. Many large vendors saw their market valuations decline during what industry observers termed the SaaSpocalypse, a period characterized by fears that generative artificial intelligence and vibe-coding could render traditional enterprise application suites obsolete. In response to this existential threat, SAP held its annual Sapphire conference in Orlando last week. The German software giant used this platform to present a counter-narrative regarding how generative AI will integrate with its vast portfolio of enterprise applications and analytics tools.

The strategy presented at the conference is defined by a distinct duality. On one hand, SAP is actively facilitating the creation of AI agents that operate using data from outside its own ecosystem. This approach suggests a willingness to engage in an open, interoperable future where data flows freely between different software providers. On the other hand, there are compelling indications that the company is simultaneously creating friction for developers who wish to build agents on third-party platforms while relying on SAP systems as their data source.

This dual approach creates a complex environment for enterprise IT leaders. The public-facing message emphasizes collaboration and extension, yet the underlying technical policies suggest a desire to retain dominance over how those extensions are managed and monetized. Understanding this tension is crucial for predicting the future of enterprise AI infrastructure.

Introducing Joule Studio 2.0

The centerpiece of SAP's announcement was Joule Studio 2.0, a new environment designed to allow developers to create and manage AI agents. A significant feature of this update is the native support for Model Context Protocol and A2A protocols. These standards are critical because they facilitate interoperability between different data sources and AI models.

By supporting these open standards, Joule Studio 2.0 can connect and collaborate with third-party tools and agents. This capability allows SAP's agentic orchestration to run across hybrid landscapes, integrating real-time data ingestion to support context-aware processes that span both SAP and non-SAP systems. The goal is to make the SAP platform a hub for AI activity rather than an isolated silo.

Muhammad Alam, SAP executive board member for product and engineering, emphasized extensibility as a core design principle during his conference remarks. He stated that users can extend any of the hundreds of out-of-the-box agents by adding tools, workflow steps, and even code through a unified experience in Joule Studio. This mechanism is explicitly designed to allow connections to non-SAP applications, acknowledging that enterprises inevitably operate across multiple vendor ecosystems.

Why does the API policy matter for AI adoption?

While the technical features of Joule Studio 2.0 promote openness, the commercial and regulatory framework surrounding them tells a different story. Gartner senior director analyst Christian Hestermann has pointed out that SAP's recently published API policy can be interpreted as an effort to control access to capabilities within the SAP platform.

This policy effectively dictates how third-party AI platforms might build agents based on SAP applications. The implications extend beyond simple data retrieval. Agents are expected to perform complex business activities and chains of operations, not just read or manipulate raw data. By controlling these interactions, SAP aims to channel who can access its systems through external AI solutions.

Hestermann noted that if SAP endorses third-party environments, it will likely charge customers extra for that privilege. This suggests a monetization strategy where openness is granted only at a premium cost. The policy serves as a gatekeeper, ensuring that while agents can roam the application estate, they do so on terms defined by SAP.

This approach creates a potential conflict with the broader industry trend toward decentralized AI development. If enterprises are forced to pay additional fees for using third-party AI platforms with their SAP data, the economic incentive shifts back toward keeping all operations within the SAP ecosystem. This dynamic raises questions about whether true interoperability is achievable without significant financial penalties.

The Anthropic Partnership and the Walled Garden

Adding another layer to this strategy is SAP's partnership with Anthropic to bring its Claude model into the SAP Business AI Platform. Analysts view this move in conjunction with the API policy, suggesting a coordinated effort to close the door to third-party AI environments while offering specific internal alternatives.

The combination of these moves indicates that SAP is constructing a walled garden around its AI capabilities. By integrating Claude directly into its platform, SAP provides a convenient option for customers who wish to avoid the complexities and potential costs of managing external agents. This strategy aligns with the broader trend of enterprise vendors seeking to become the locus of control for agentic AI.

Competitors like Salesforce, Oracle, and ServiceNow are pursuing similar strategies, positioning their own technologies as the central hub for AI workflows across mixed vendor environments. The technical question of whether these protocols work has largely been resolved. The remaining challenge is economic: who pays for runtime, who governs the agent, and whose roadmap dictates future development.

How does this impact enterprise contract negotiations?

The strategic moves by SAP have profound implications for long-term enterprise IT planning. Faram Medhora, principal analyst at Forrester, argues that the choice of platform for building cross-vendor agents is not merely a technical decision but a commercial one with decade-long consequences.

Medhora warns that the platform an enterprise selects to build its first cross-vendor agent will likely price its entire AI estate for the next ten years. This reality frames current decisions as early stages of contract negotiations that will play out over the coming decades. The technical feasibility of connecting SAP agents to Salesforce or other systems is irrelevant if the economic terms are unfavorable.

Large enterprise IT departments are under pressure to execute an agentic AI strategy quickly. However, rushing into a platform without considering the long-term commercial implications can lead to vendor lock-in and inflated costs. The current landscape requires careful evaluation of who owns the audit trail and who controls the governance of these autonomous systems.

SAP wants to be that choice among its customers. This makes sense if an enterprise relies heavily on SAP applications, but most large organizations use a mix of vendors. The commercial reality presents challenges that technical interoperability alone cannot solve. Enterprises must navigate this complex terrain by balancing immediate operational needs with long-term strategic autonomy.

What are the practical takeaways for IT leaders?

For IT leaders, the announcement of Joule Studio 2.0 signals a shift in how SAP intends to participate in the AI economy. The support for open protocols like MCP and A2A is genuine and valuable, allowing for greater flexibility in agent design. However, this openness comes with conditions that must be carefully scrutinized.

Enterprises should evaluate the total cost of ownership when integrating third-party agents with SAP systems. The API policy may introduce hidden costs or restrictions that could undermine the benefits of interoperability. It is essential to understand the governance model and audit trail ownership before committing to a specific platform for AI development.

The partnership with Anthropic offers a streamlined path for those willing to stay within the SAP ecosystem, but it also highlights the vendor's desire to keep customers close. IT leaders must weigh the convenience of integrated solutions against the potential loss of flexibility and control over their AI infrastructure.

Ultimately, the decision will depend on the specific mix of applications an enterprise uses and its long-term strategic goals. While technology allows for cross-vendor collaboration, the economic framework determines whether such collaboration is sustainable. The coming years will likely see intense competition as vendors vie to become the central hub for agentic AI in the enterprise world.