SAP FUE for AI Agents

AI license matrix for S/4 Hana agents

SAP license management was once considered a domain of complex rules and opaque metrics. Over the past three years, it has become an even denser jungle, with many SAP customers finding it difficult to navigate. SAP's cloud strategy and the implementation of Full Use Equivalent (FUE) as a new licensing model are at the heart of this development.

Current SAP license metrics are based on past experience and requirements. They make no mention of new technologies, such as agent-based AI. Excluding AI systems for optimization, orchestration, and automation could be disastrous for SAP. From this perspective, AI appears to threaten a traditional ERP system.

A central element of SAP's cloud strategy, particularly in the context of RISE with SAP, is the FUE concept. FUE is the conversion key that determines how existing on-prem licenses are transformed into cloud subscriptions. It is a complex set of rules with detailed instructions on how to make the transition from the traditional on-prem model to the cloud.

The FUE metric aims to simplify license management by converting previous usage-based user categories into authorization-based equivalents. SAP defines specific conversion factors for different S/4 Cloud usage types. AI agents are not included in this calculation. SAP assumes that agent-based AI is irrelevant to ERP.

FUE for AI agents

One FUE (Full Use Equivalent) corresponds to one S/4 for Advanced Use (comparable to Professional User in the on-prem area). Approximately two FUEs are required for S/4 Cloud Development Access. Interestingly, applications such as Key User Extensibility, Side-by-Side Extensibility on SAP BTP, administrative SAP tasks/users, and SAP Fiori extensions do not require a separate Development Access license.

Is this strategy or blindness? It is unclear how AI agents based on the SAP BTP (Business Technology Platform) will be treated. If AI agents on the BTP were to operate in a legal vacuum, SAP could lose a lot of license revenue quickly. Artificial intelligence has an inherent destructive nature. Why not use Lovable and a few AI agents to create a new, composable ERP system based on the SAP BTP and BDC? Traditional ABAP ERP and AI could put SAP in a tight spot in the cloud. With Lovable and other AI tools, large numbers of AI agents could be created quickly.

DeepL Agent is a new autonomous AI agent that supports knowledge workers by taking over repetitive, time-intensive tasks in various areas, freeing up capacity and boosting productivity. The agent is designed to work entirely within each user's digital environment, responding to natural language commands to safely and independently perform complex workflows. It uses virtual versions of standard tools, such as a keyboard, browser, and mouse, to act on behalf of the user via existing interfaces. The agent can handle almost any task that a human can perform using computer systems. It moves seamlessly between users' tools and workflows, continuously improving its performance over time.

Agent-based ERP Full Use

In the future, it is conceivable that a complex SAP S/4 system will be controlled by a few AI agents, while numerous employees will manage the agents and no longer interact with the ERP system directly. An FUE license would cost a fraction of the current price.

History repeats itself: my first SAP R/3 system ran on an IBM RS/6000 with the Unix derivative AIX. The operating system recognized only three users: the administrator (root), the database administrator, and the R/3 system. The AIX license was inexpensive. However, over 100 users worked on the R/3 system. However, AIX was unaware of this! While SAP benefited from lucrative licensing, IBM received no added value from the service.

“Imagine a highly efficient assistant who understands your needs and can help with any task, such as analyzing a report or managing internal invoices. It can perform these tasks with simple instructions, just as you would give to a colleague. This is exactly what DeepL Agent does. It is designed to understand you, so you can simply describe the desired actions and it does the rest,” says Stefan Mesken, Chief Scientist at DeepL. “We designed the agent to learn from previous interactions, so its support becomes more personalized and tailored to each user's requirements and daily workflows over time.”

If agent-based AI takes over operational ERP, then SAP's business model is in grave danger. Not only could a DeepL AI agent customize SAP S/4 Hana according to its own ideas, but it could also take over the work of hundreds of users. Numerous professional user licenses would then become obsolete. The SAP license system would collapse like a house of cards with AI agents. Ultimately, the AI agents will customize their own open-source ERP based on their S/4 experience.