Artificial SAP Chaos and AI Disaster

From the beginning, SAP lacked an AI roadmap. The beginning was promising, but the AI strategy failed, leaving SAP one step away from the artificial abyss. The technical basis for ERP intelligence is anchored in the SAP core, particularly in Hana. The Predictive Analytics Library (PAL). Hana PAL is not a generative AI tool; rather, it is a library of machine learning (ML) and data mining algorithms. These algorithms, such as decision trees and K-means, work directly on application data, which is essential for tasks such as predictive analytics. This approach is primarily used for classification, while the current revolution is taking place in large language models (LLMs) and generative AI.

The situation surrounding the introduction of AI in the SAP environment is akin to modifying the complex cybernetics of a living organism. The regulars at my SAP club successfully operate their ERP systems, and I can already hear them saying, "never change a running system."

Although new components, such as LLMs and AI agents, are being delivered and promise tremendous speed, SAP customers must ensure that the new technology runs stably on the old infrastructure (ERP/ECC 6.0). They must also keep their hand on the wheel (BTP, including the GenAI hub) to prevent the AI power from tearing the entire ship apart.

The term "cybernetics" is derived from the ancient Greek word for the art of steering! SAP board member Thomas Saueressig spoke of a Frankenstein-like architecture in connection with AI and on-prem ERP. In his opinion, this composite monster is a warning against composable ERP with Hana, S/4, Salesforce, Workday, and ServiceNow. Similarly, his fellow board member Muhammad Alam outlined the danger of an ERP patchwork quilt. SAP wants exclusive, end-to-end processes in the ERP cloud.

AI has the potential to transform ERP processes that have evolved over decades and manage automation. BTP GenAI Hub, LLMs, and AI agents provide the technical capability to surpass traditional Hana PAL predictions. However, we customers face a bipolar challenge.

On the one hand, there is the need for technical adaptation in order to remain competitive; on the other hand, there is a lack of transparency, and a lack of strategic leadership from SAP. Additionally, integration, licensing, and governance costs are high and difficult to calculate. Until SAP provides clarity and reliability in this area, using AI in the ERP environment will be a balancing act between innovation and risk management.

The discourse on AI in the SAP environment has long since left the theoretical phase and become a central—albeit critically viewed—driving force behind digital transformation. From the perspective of SAP customers, AI is not an isolated IT project but rather a profound change intended to make processes more efficient, faster, and higher quality.

P's AI chaos theory reaches its preliminary climax in RPT-1. According to my friends and informants, RPT-1 is supposed to be a narrow-gauge LLM. After Hana PAL, SAP wants to try its hand at machine learning. AI rules will be derived from anonymous data from customers and used by Joule, which consists of AI assistants and agents. In fact, the planned AI system, RPT-1, is not a narrow-gauge LLM but, according to SAP, merely a "Relational Pretrained Transformer."

No one would tell me if RPT-1 uses Hana objects, vector engines, or graph engines. The most visible result of the AI strategy is Joule, which proactively delivers context-related insights and has been integrated into the cloud portfolio. And RPT-1? PAL algorithms transform anonymized customer data to pre-train AI agents according to the relational database principle of Hana PAL—that's AI chaos theory.