95% of students already use AI in class, but only 1 in 4 teachers feels ready to guide them

By Zendoric · September 29, 2026.

The debate running through this article is simple and uncomfortable at the same time: students and teachers already interact routinely with artificial intelligence tools, but a large share of the teaching staff does not feel adequately prepared to guide that use with pedagogical judgment. The source does not give exact figures, but the framing precisely describes the mismatch that defines this phase of adoption: the technology has entered the classroom far faster than the training of those who must accompany it.

The shift the article proposes points in a recognizable direction: education must teach students to think, decide, verify information and solve problems with their own judgment. It is not about banning AI in the classroom —that battle is already lost— nor about adopting it without question. It is about redesigning which capabilities matter when part of the routine intellectual work is already done by the machine: critical thinking, source verification, and communication and argumentation skills. Broadly, the debate sits not only in the academic sphere but also in that of the well-being of a generation growing up with these tools woven into daily life, not as an occasional novelty.

This connects directly with something we have been observing in our tracking of AI's impact on employment by sector: in education, the winner is the teacher who learns to orchestrate AI as a tool for augmented tutoring, and the loser is the one who merely transmits content that any model already reproduces better and faster. The lack of teacher preparation is not an anecdotal curiosity: it is the practical manifestation of that thesis. If a substantial share of teachers do not feel equipped to guide AI use, the transition will not be smooth, and the risk of a widening gap between schools with the resources to train their staff and those without them is real and deserves to be taken seriously, not minimized with easy optimism.

In the short term, the problem is not hypothetical: students delegating to AI the very tasks meant to build underlying capabilities —reasoning, arguing, verifying— without any consolidated pedagogical framework yet in place to prevent it. It is exactly the kind of mismatch to be expected from any general-purpose technology adopted faster than those who must supervise it can be trained: it happened with the internet and with search engines, and it is happening now on a larger scale and at greater speed.

In the longer term, however, the direction of change looks right to us, even if the journey is bumpy. If schools manage to shift the focus from memorization toward judgment, verification and self-regulation —the capabilities a language model does not replace, but rather demands be strengthened in order to use it well—, they will be cultivating precisely the kind of thinking that will remain valuable in an economy where AI takes on ever more routine tasks. That, ultimately, is the long-term bet on abundance: an education that frees up time and attention to think better, not just to produce faster. But that bet only pays off if the teacher-training gap this debate puts on the table is closed, and that is a problem of investment and political will today, not in the next decade.