AI agents in Composable ERP

Agentic AI: from supply chain management to financial closing.

SAP CEO Christian Klein's strategic justification for the „all-in on AI“ change of course is compelling: artificial intelligence absolutely requires the proprietary business data and deep process logic of an ERP system, as isolated language models without this specific business context are prone to dangerous hallucinations and cannot deliver reliable business added value.

By using the AI co-pilot SAP Joule and collaborative multi-agent systems, SAP promises users huge productivity boosts and cost savings of up to 30 percent, as repetitive routine tasks will be almost completely taken over by machines in the future.

SAP Joule for Consultants

A central, organizational and technical tool of this AI offensive is „SAP Joule for Consultants“, a specialized assistant that provides consultants and key users with the concentrated knowledge from SAP Notes, official documentation and best practices in a targeted manner in order to massively accelerate the often tough, expensive and painful S/4 transformation projects.

The generative AI Hub on the Business Technology Platform (SAP BTP) forms the architectural foundation for the holistic agentic vision. SAP is not positioning itself as a developer of its own universal base models, but is acting as a broker that provides an orchestration layer through which existing customers can access powerful large language models (LLMs) from partners and hyperscalers such as OpenAI, Anthropic, Amazon and Meta. With the help of tools such as SAP Joule Studio, developers in user companies will be able to construct their own customized AI agents that securely link external AI knowledge with sensitive internal SAP data.

Toxic mix of technology and marketing

However, if you look at these cloudy SAP promises from a critical distance, the AI strategy for many existing SAP customers often turns out to be a toxic mixture of unfinished technology and organization as well as a forced sales corset. In the reality of IT departments, the devastating verdict of many developers and consultants is that the loud SAP marketing is miles ahead of actual operational maturity and that SAP's AI technology still lacks essential skills for error-free everyday use.

Autonomous AI agents that independently carry out business-critical bookings or orders require excellent data quality and seamless governance, which simply does not yet exist in the historically grown, fragmented data landscapes of many existing customers. This skepticism is also underpinned by the latest investment report from the German-speaking SAP User Group (DSAG), which ruthlessly unmasks SAP's AI dreams: When companies implement productive AI use cases in their business processes today, an alarming 77 percent resort to non-SAP solutions, while only a marginal three percent use SAP's original AI tools in their day-to-day business.

Rise with SAP and the AI sugar coating

However, the most existential danger for sovereign existing customers lies in the commercial blackmail by the SAP Executive Board, which is deliberately misusing the AI sugar-coating as a lever to force hesitant companies into expensive public cloud contracts such as Rise with SAP or Grow with SAP. Those who continue to operate their systems on-prem for compliance or data sovereignty reasons will simply be excluded from essential AI innovations and the Joule assistant.

In this context, existing SAP customers fear not only an inevitable vendor lock-in and completely non-transparent license costs for indirect AI data usage, but also the creation of an unmanageable „Frankenstein architecture“ in which AI agents from a wide variety of software manufacturers collide with each other in an unregulated, chaotic patchwork in an uncontrolled manner.

Ultimately, the enlightened user must recognize that Christian Klein is primarily acting as a driver of the financial markets in the field of artificial intelligence, who must necessarily deliver a growth story for his own share price instead of guaranteeing his customers an economically secure, ethically mature and, above all, business model-independent AI evolution at eye level.

Just as the artificial creature Frankenstein was assembled from numerous body parts of deceased people, SAP CEO Christian Klein's „All-in on AI“ offering is also currently appearing. It is no coincidence that his fellow board member Thomas Saueressig warned against an ERP Frankenstein architecture six months ago. Can AI agents tame the Frankenstein monster?

Under the loud mantra „All-in on AI“, SAP CEO Christian Klein is staging a far-reaching organizational, technical and licensing realignment in which AI is to fundamentally redefine and revolutionize traditional ERP business software. But there is a great danger that a composable ERP will become a Frankenstein architecture AI.