From Generative AI to Agentic AI: What Challenges Lie Ahead for Schools?

Agentic AI: When Machines Become Proactive

While schools are only beginning to integrate generative AI tools, a recent study (available in French for now) by the International Observatory on the Societal Impacts of AI and Digital Technologies (OBVIA) points to the rapid emergence of agentic AI systems. According to the document, these systems are defined as “autonomous systems capable of making decisions independently through self-learning capabilities, without regular human supervision.” In education, this could include automated tutors capable of structuring knowledge and evaluating learners autonomously.

Unlike the reactive tools that educators are already familiar with, agentic AI is proactive: it can plan, coordinate, and execute complex tasks in pursuit of self-defined goals. According to the authors, this increased autonomy “intensifies issues that already exist” and raises serious concerns about human oversight.

The Training Paradox: 73% of Students Left on Their Own

One of the study’s most striking findings concerns the still “incomplete” institutional integration of AI in schools. Research cited in the report reveals that 73% of students “reported receiving no formal training on the use of generative AI as part of their educational program.”

This lack of guidance is creating new educational inequalities. As the study points out, the digital divide is no longer simply about technical access, but also about the ability to formulate effective prompts and assess the reliability of generated content. Without structured support, AI tends to benefit primarily students who already possess strong analytical and self-regulation skills.

Cognitive Risks and “Cognitive Substitution”

One of the report’s most significant concerns for educators involves the impact on learning itself. The report warns against what it describes as “cognitive substitution.” While AI may improve the formal quality of assignments, researchers argue that it can also become a “cognitive shortcut” that reduces intellectual effort and weakens retention.

The document strongly reiterates one fundamental principle: “Completing a task successfully using AI does not automatically lead to learning.” According to Obvia, the gap between “false mastery” and genuine learning depends entirely on whether AI use is guided by clear pedagogical principles.

Without a strong instructional framework, AI can become a form of “cognitive debt” that weakens students’ critical thinking skills—particularly among younger learners, who may conceal gaps in understanding behind polished, machine-generated work.

Rethinking Assessment Urgently

Faced with AI agents capable of completing entire courses, educational institutions must rethink how learning is assessed. According to the study, approximately 71% of students already perceive AI as a threat to academic integrity.

To address this challenge, the report recommends assessment formats that are more resistant to undisclosed AI use, including:

• oral examinations and in-class assessments;
• portfolios and learning journals;
• evaluation of the learning process rather than the final written output.

Preserving Professional Autonomy

Finally, the study examines the teacher’s role in an increasingly automated educational landscape. While AI can support teaching and instructional design, its widespread use could lead to an “implicit normalization of pedagogical practices” shaped by the technical choices of major technology companies.

For educational consultants, the challenge is ensuring that technology remains secondary to educators’ professional judgment and expertise. As emphasized in the study’s education section, AI should not become an end in itself, but rather a tool for developing learners’ agency, critical thinking, and autonomy.

An English version of the Obvia 2026 report is expected soon.