Orientation paper criticizes current state of AI in the education system

When it comes to implementing AI competencies and AI in German schools, structural inequality is the most fundamental problem in the education system. This or something similar is how the criticisms found in the current orientation paper “AI-related School Leadership Training in Germany” by Forum Bildung Digitalisierung can be broadly summarized. To put it more sharply: Regarding AI, the German education system is acting in its usual federally headless manner, even though it is apparently structurally dependent on a functioning trickle-down effect. But if not much comes from above, it's a matter of luck if anything trickles down at all.

The orientation paper therefore calls for a nationwide AI strategy as well as more support, resources, and at the same time less bureaucracy, to actually achieve educational equity regarding AI: students should have the opportunity nationwide to achieve similar competencies and levels of competence. A goal that is often mentioned by educational politicians.

If one believes the anonymized voices from various actors in the education system compiled in the orientation paper, Germany is currently far from this. Although the text primarily concerns AI-related qualification of school leaders, as this is dependent on overarching structures, the spotlight ultimately falls on state institutes, which usually offer training, school authorities, school supervision, and school policy at federal and state levels.

Everyone cooks their AI soup

Employees of state institutes have therefore stated to the authors of the paper that the different personnel and financial resources of the states in the field of education alone affect the implementation of AI solutions and AI competencies: “From the perspective of some interviewees, this unequal starting position leads to AI-related innovations and competencies not being able to develop comprehensively and equally in the education system.” But even if personnel and finances were available, the different federal states are not moving in the same direction and often do not even speak the same language to be able to cooperate at all - if they wanted to.

For example, a person from a state institute reports that they repeatedly find in the institute “that the materials developed in other federal states are partly not applicable here at all because different terms are used.” In addition, there is the legal situation. The orientation paper states: “In Germany, there is no nationwide strategy that regulates uniform legal frameworks for the use of AI in schools. In the federal education system, each federal state develops its own guidelines. This leads to different requirements regarding the application of AI tools.”

The different legal requirements also make the work of private providers of AI training difficult. They criticize that they “cannot close existing training gaps, as different legal frameworks, for example regarding the admissibility of certain AI applications, make it difficult to design training uniformly nationwide.”

Do we have a plan, and if so, how many?

Meanwhile, school leaders are being confronted with the demand that schools are responsible for preparing learners for new societal conditions “so that they can act autonomously and remain capable of action in a society and working world shaped by AI.” For this purpose, they and their teachers should best apply the recently expanded Dagstuhl triangle. In its original form, it includes the guiding question, “How does it work?” (Understanding), “How do I use it?” (Applying), and “How does it affect?” (Reflecting). For AI school development, Alles, Falck, Flick & Schulz developed a separate AI competence model in 2025, which, in addition to the three Dagstuhl perspectives, adds a fourth area: "Co-creating".

School leaders should therefore acquire the ability to “actively initiate transformation processes with and through AI, open up design spaces for the faculty and student body, promote a reflexive innovation culture, and further develop the school as a learning organization.” For them to be able to do this, however, they must first be able to understand, apply, and reflect on AI, and here too the system does not deliver what it should. According to the orientation paper, about half of the federal states offer corresponding AI training courses through their state institutes. These training courses often include the Dagstuhl triangle but not the extension recommended by the orientation paper. And when training courses are offered, they are usually not mandatory and often last only a few hours. According to the paper, the few training courses over a longer period of time are considered more sensible.

Can school leaders rely on other help or structures when it comes to AI and implementation? The paper describes the situation as follows: “In the support system, including school supervision, which accompanies, controls, and steers quality development in the school system, there is often a lack of AI competencies and corresponding training offers, according to the interviewees, which makes everyday support for school leaders in this area, beyond training, difficult.” Crucial actors in the education system often lack AI competencies themselves. A person from a state institute expresses this more drastically: “Then we have school supervision, which is supposed to support, advise, and steer the entire system. But they are precisely the furthest removed from current developments because there have been no explicit training offers in this field so far. How am I supposed to advise and support schools on issues of students' network competence if I cannot deal with it myself? School supervision also has no direct access to AI applications so far.”

The authors of the paper classify this situation as a “risk” and refer to examples from abroad, where the respective school administrations are considered “a central actor for the use and governance of AI in the education system.” For example, the “AI Guidebook of the Chicago Public Schools” provides a clear framework for the responsible use of generative AI in education. It explains basic concepts, describes useful use cases, and defines binding rules on data protection, fairness, and academic integrity. However, this example also shows that internationally there is a lot of fragmentation, coexistence, and incompatibility.

The paper must therefore also admit that Germany is not entirely alone in its current situation. It does so indirectly by stating that national strategies for the legally compliant use of AI in educational institutions have so far hardly been developed; Estonia or some Asian countries are considered few exceptions. And it summarizes: “The lack of overarching political guidelines makes practical implementation difficult not only abroad but also the development of consistent training offers for school leaders.” Consistent and overarching guidelines, strategies, and support remain the key words here as well.

(kbe)