European Trade Union Committee for Education

Across Europe and globally, education systems are trying to understand how generative artificial intelligence (AI) will affect teaching, learning, assessment, and educational quality. While much of the debate has focused on technology, regulation, or academic integrity, AI in education is fundamentally a pedagogical and professional challenge. It affects core dimensions of educational practice, including writing processes, assessment, student agency, and the role of teachers themselves.

In Denmark, our response to AI has therefore not primarily been based on rapid regulation or top-down implementation. Instead, the Ministry of Education and the Danish upper secondary teachers’ union, GL, have supported four national pilot projects intended to generate practical and research-informed knowledge about AI in education.

Officially, the projects investigate how AI affects writing, learning, and assessment. In practice, however, they have evolved into something broader: a collaborative process where teachers, school leaders, researchers, policymakers, and unions jointly explore how AI should shape educational practice. The projects therefore function simultaneously as AI pilots, professional learning initiatives, and policy-development processes.

The projects emerged in response to growing uncertainty among teachers and schools. During workshops organised by the Danish Agency for Education and Quality (STUK) in late 2024, teachers raised concerns about written assignments, assessment practices, examinations, and students’ writing competencies in an AI-rich environment. A central issue was the contradiction between classroom practice and examination systems: AI was increasingly present in students’ everyday learning processes while simultaneously being prohibited in many forms of assessment.

In response, the Ministry launched four pilot initiatives focusing on different aspects of AI and educational practice: AI in interdisciplinary written assignments, AI-supported synopsis work in Social Science, portfolio-based writing with and without AI in Danish, and an assessment experiment combining written and oral grading in science subjects. Together, the initiatives function as a national laboratory for examining how AI affects learning, writing, and assessment.

Importantly, the projects were designed not as technology implementation programmes, but as exploratory learning initiatives intended to generate knowledge before broader reforms or regulation are introduced. In an international context where AI policies are often shaped by technological hype, market pressures, or political urgency, we have instead attempted to create space for classroom experience, professional judgement, and research evidence to interact.

What makes these projects distinctive is the way they combine classroom experimentation, collaborative reflection, research, and policymaking within the same process. Rather than positioning teachers as passive recipients of externally developed AI strategies, the projects position teachers as active producers of professional knowledge. More than 400 teachers voluntarily signed up to participate, representing perspectives ranging from highly enthusiastic to strongly sceptical views on AI. This diversity has helped ensure that the projects function as spaces for professional deliberation rather than technology advocacy alone.

One of the most important dimensions of the projects has been the role played by social dialogue. At a crucial stage in the process, GL argued that the pilot projects should include systematic follow-up research rather than merely administrative evaluation. This intervention fundamentally changed the character of the initiatives. The projects became not only pilot experiments, but evidence-generating professional learning processes.

This is an important example of teacher agency shaping educational policy. Rather than opposing innovation, the union contributed pedagogical expertise, professional legitimacy, and classroom-based knowledge. In our experience, this demonstrates how unions can function not only as defenders of working conditions, but also as professional knowledge organisations and co-creators of educational policy.

A central strength of the projects has been the close integration between research, professional dialogue, and policy development. Three of the four initiatives conclude in summer 2026, while the long-term portfolio project continues for an additional two years. As part of the conclusion of the three shorter projects, three national evaluation conferences are currently being planned.

These conferences are designed as collective professional learning arenas where research findings will be discussed alongside pedagogical experiences and dilemmas from practice. Teachers will receive new knowledge and inspiration while also engaging in collective reflection about the implications of AI for educational practice.

Most importantly, participating teachers will formulate recommendations regarding the future use of AI within each project area. These recommendations will subsequently be distributed nationally by the Ministry of Education. The conferences therefore function simultaneously as professional development, democratic dialogue, knowledge dissemination, and policy formation. Teachers are not positioned merely as respondents within reform processes; they actively participate in interpreting evidence and shaping future recommendations.

Ultimately, our work with these pilot projects suggests that the governance of AI in education is not merely a technical question, but a democratic one. In many contexts, educational technology policy risks being shaped primarily by commercial interests or technological solutionism. Our approach instead seeks to build on professional trust, social dialogue, and collective knowledge production.

The projects demonstrate how technological change can become a catalyst for collaborative professional learning rather than a purely technical reform process. Through social dialogue, teacher agency, union involvement, and research-informed collaboration, we have attempted to create an integrated model connecting classroom experimentation, professional learning, evidence generation, and educational policymaking.

Our experience so far suggests a broader lesson for education systems internationally: if AI is to support quality education, teachers and their professional organisations cannot simply adapt to technological change. They must actively participate in governing it.