A number that nobody expected.
According to model calculations by the German Kultusministerkonferenz The Standing Conference of the Ministers of Education and Cultural Affairs (KMK) estimates the number of teachers in the German education system at between 49,000 and 76,000. By 2032, an additional 600,000 children will enter the education system. The situation is similar in other countries, such as Austria. Teachers' forums are up in arms, as many still remember the recommendation from the 2000s to avoid becoming teachers. The reason given was that there were too many teachers and, in the long run, too few positions. Letters to this effect were even sent by the Ministry of Education to individual households and high school graduates at the time.
The essential question today – twenty years later – is no longer: Do we need more teachers? That question can probably be answered unequivocally with "yes" based on the figures above. The question is: What do we do with the teachers we already have?
The OECD measured in its latest TALIS surveyHow teachers in 48 countries spend their working hours. The result: On average, 40 percent of working hours are not spent on direct teaching. They include documentation, reports, compliance, preparation and follow-up, differentiation, parent communication, and corrections.
What really helps: support systems before, not during, the class.
More precisely, a teacher's "core business" is teaching. Let's stick with that. A differentiated lesson for a class requires approximately:
-
-
Individual research and material selection, adapted to the class and situation: 20-30 minutes
-
-
-
Adaptation to curriculum and competency levels: Should already be in place.
-
-
-
Differentiation would be desirable. This (let's see) and variations for higher-achieving and lower-achieving students: 10-40 minutes
-
-
-
Creating your own new tasks and exercises: 20-30 minutes
-
-
-
Multilingualism for classes with German as a second language: an additional 20-30 minutes
-
From this list, one thing will surely be clear to the esteemed reader: if one wishes, a single lesson can take up an entire day to prepare. And: a single well-prepared lesson realistically takes between one and three hours. A week with 20 lessons results in 20-60 hours of preparation—if everything is thoroughly planned and documented. Even if this work is attempted during the holidays. In August, one doesn't yet know what the particular interests of class 3B will be in November.
Assuming 20 teaching hours, 3-6 hours of general administrative and organizational work (estimated), and a lower estimated end to preparation, plus approximately 10-20 hours of general marking work per week (homework, small checks, feedback), even at the lower to middle end of the estimate, one arrives at 20 + 3 + 10 + 20(?) = 53 hours per week. These hours are neither clearly measurable nor clearly documentable.
And those that go unseen.
What AI support practically enables
Studies from 2025 and 2026 show a clear finding: AI tools save teachers who use them regularly, for example 5.9 hours per week minimum. A study with 336 college lecturers in the USA showed a time saving of 6.5 hours. These hours can provide valuable support for the above workload. Adding to the above list reveals practical applications:
Individualized research and content selection tailored to the class and situation – the AI conducts the research when a specific topic is specified – the teacher can also indicate which topics the class currently favors or what would be particularly useful. The textbook serves as the foundation; the teacher adapts the material to the situation.
Adaptation to curriculum and competency levels: A detailed plan for reference and supplementation is available with AI support.
Differentiation by language and competence levels integrated by AI
Task creation is then fully implemented, can be supplemented upon request, and supports multilingualism.
McKinsey estimates that 20 to 40 percent of some teachers' tasks could be automated. Not the teaching itself. Not the relationship with the class. Not observation and understanding. Not guidance and the human element.
AI for teachers makes sense.
Die Scandinavian countries have just demonstratedthat digital tools in the classroom can worsen learning. Student tools are not the answer to the teacher shortage. The answer is a completely different kind of AI—one designed for the hours before class. The hours after class. The hours when teachers distill the material from the curriculum, the class, and the topic, material that students then hold in their hands.
This model has a name that is currently becoming established: Teacher-First-Modell.
What Teacher-First actually means
In the Teacher-First model, the teacher remains in the decision-making and design role throughout the learning process. AI provides pre-prepared building blocks that the teacher selects, adapts, combines, or discards. No AI is used in the classroom—that's where human learning takes place.
Specifically, this means:
What AI does: Research topics. Differentiate materials in seconds. Create quizzes. Generate activities. Deliver multilingual versions. Establish curriculum links. Incorporate Bloom's Daylight Saving Time.
What the teacher does: Decide if the material is suitable. Adapt it if necessary. Be present in the classroom. Observe. Explain. Build relationships. Evaluate.
What students see is not AI. They see their teacher with a finished worksheet. What they don't see is that this worksheet was created in five minutes instead of 90.
The turning point lies in the method.
OECD data and burnout research lead to the same conclusion: What truly supports teachers is not a new teaching concept. It is relief from the burdens outside of the classroom.
The right question is not: How do I replace what is currently happening in schools? But rather: How do I relieve the burden on teachers without interfering with what students need? Support that is efficient and, above all, substantive for the work before and after lessons.
Whoever answers this question correctly will build the bridge that has been called for in every education report for the past two years – and can be delivered in a better way.
Sources: OECD TALIS 2024, Gallup/WFF 2025 on AI time savings of 5.9 hours per week, McKinsey 2024 on the automation rate of administrative tasks, Standing Conference of the Ministers of Education and Cultural Affairs (KMK) model calculations 2024-2026, Forsa survey Deutsche Telekom Foundation March 2026, German School Barometer 2025 (Robert Bosch Foundation), RAND 2024 on burnout factors, ÖLI-UG working time model, Mark Rackles/Deutsche Telekom Foundation 2023 on the teaching load model, Standing Scientific Commission of the KMK 2023, Arkansas State University study 2026 on AI time savings (336 lecturers, 6.5 hours/week), Open Educat research overview “Reducing Teacher Burnout With AI” 2026.
