AI Agents in SAP, with Risk Check

SAP is one of the pioneers in the development of AI solutions for businesses. These technologies are designed to accelerate, automate, and scale business processes across a wide range of areas. However, there is one major catch: SAP’s native AI tools are only fully available to companies using the latest versions of the software.

Most companies don’t do this—and for good reason. SAP systems have often evolved over decades and are deeply integrated into business processes. Bringing all environments up to date is correspondingly time-consuming. Nevertheless, companies that are not currently considering an upgrade do not have to do without AI. Instead of using SAP’s native AI functions, they can develop their own AI agents and integrate them into their SAP landscape as extensions.

AI Agents: Opportunities and Risks

This approach opens up new possibilities, but it also presents challenges. The greatest risk is that external AI agents will create additional technical debt and introduce governance risks that would be easier to avoid with native SAP solutions. With its Clean Core strategy, SAP aims to keep customer-specific customizations as far away from the platform’s core as possible. This is intended to simplify maintenance, innovation, and future upgrades. However, companies that deeply integrate third-party agents into their SAP systems run the risk of undermining this very principle.

In the long term, this could result in future SAP upgrades becoming even more complex than before. Added to this is the risk that SAP’s governance, compliance, security, and data protection mechanisms will not integrate seamlessly with external agents because their architecture and functionality do not fully align with SAP’s specifications. At first glance, the decision therefore seems difficult: Should companies deploy custom AI agents and accept the associated risks? Or should they forgo the benefits of AI in their ERP system to avoid technical debt and governance issues?

Using AI in a Controlled Manner

In fact, this is a false dichotomy. Companies can deploy individual AI agents in SAP without allowing risks and technical debt to spiral out of control.

The key lies in what are known as AI harnesses—that is, guardrails and control mechanisms that ensure the secure use of AI agents within SAP. In practice, this usually involves a combination of native SAP functions and additional third-party control tools. For example, policies within the SAP Business Technology Platform (BTP) can be used to specifically restrict data access by AI agents. At the same time, identity and access management solutions from hyperscalers such as Microsoft Entra can be used to register and authenticate agents.

The result is an architecture that effectively reduces governance risks without creating new technical legacy issues.

The Path to the Future

In the long term, modernizing the SAP landscape and leveraging the vendor’s native AI capabilities remain the most sensible way to comprehensively „agentify“ business processes. Until this step becomes a realistic option for many companies, however, a pragmatic approach is advisable: the use of individual AI agents in combination with the SAP Business Technology Platform. This allows companies to benefit from modern AI capabilities today without losing sight of security, compliance, and governance. (Source: Lemongrass)