Education-Specific

On Monday, May 11, the Council of the European Union met in Brussels and approved its first formal conclusions on the role of teachers in the era of artificial intelligence. The text is short by Brussels standards, and most of it reads in the calm institutional cadence that European policy tends to favor.¹ It is, by the Council's own framing, the first time the relationship between AI and teaching has been formally taken up in EU education policy.²

Reading it slowly is worth the time. Threaded through the conventional language about literacy, equity, and well-being is a phrase that does quiet work. The Council calls on member states to "promote education-specific AI tools."¹ Not AI tools. Not any AI tools. Education-specific.

It is the kind of phrase a careful committee writes when it has noticed something it does not yet want to name out loud.

What "Specific" Is Saying

For the better part of three years, the market has assumed a particular shape. A small number of general-purpose chatbots, trained on the open internet for the open marketplace, have been the default AI experience in nearly every classroom. Districts have written policies around them. Teachers have improvised. Students have used them as research partner, ghostwriter, study buddy, and crutch, sometimes all in the same evening.

The Council's word, "specific," is a quiet acknowledgment that this default is not what education has been asking for. A chatbot built to serve a billion consumers does not know which sentence the student tried to write and deleted. It does not know that the teacher's goal this week is to get the class to notice their own assumptions. It is, by design, oriented toward producing fluent output. The teacher's job is oriented toward producing learning. These are not the same thing.

The OECD made the same observation in January. Its Digital Education Outlook 2026 concluded that generative AI can support learning when guided by clear teaching principles, but if used without pedagogical guidance, outsourcing tasks to AI simply enhances performance with no real learning gains.³ Performance is not learning. The Council, four months later, has translated that finding into a policy verb. Promote tools that are specific.

The Teacher in the Loop

The second half of the Council's conclusions makes the first half operable. Teachers, the document insists, are not simply users of AI. They are guides, mentors and critical thinkers who help students navigate a complex digital world.² More to the point, the Council asks member states to ensure that teachers have an opportunity to contribute to the design and evaluation of AI tools used in education.¹

This is the line any vendor selling into European schools should read twice. It is the difference between a market in which education is a deployment target and a market in which education is the design partner. The first market is what we have. The second is what is being asked for.

There is good evidence the second produces better tools. In a three-study investigation across middle schools in Pennsylvania and California, led by Kenneth Koedinger and colleagues at Carnegie Mellon, a hybrid program that paired AI tutors with human tutors produced significant gains in learning, with the largest effects among students who had been furthest behind.⁴ The teacher did not disappear. She became more legible to herself.

What Visibility Has to Do With It

An education-specific AI tool, in the spirit the Council seems to mean, is one designed around what teachers actually need to see. Not output, but trajectory. Not the polished draft, but the small ordinary acts that produced it. The pauses. The rewrites. The question a student asked the AI at 9:47 on a Tuesday night and then thought through and answered for themselves.

This is the bet we have been making at Koan. When the student writes inside a workspace that captures her drafts, her conversations with Aidan, her revisions after feedback, the AI stops being a black box that returns essays and starts being part of a record the teacher can read. The Council's two requirements, that tools be specific and that teachers help shape them, point at the same building. Visibility is the floor on which both stand.

The First Time They Said It Out Loud

The most interesting thing about the May 11 conclusions is not what they propose. It is that they were the first formal EU statement on the relationship between AI and teaching at all.² For two years, the conversation has lived in vendor marketing, district pilots, and frantic op-eds. On Monday, twenty-seven countries' education ministers sat down in Brussels and, calmly, in a sentence, named the shape of the tool they want.

The next decade of educational technology will be built or refused on that sentence. A tool built for everyone is not a tool built for a classroom. A classroom needs to see its own work.

If the question is not whether to use AI in schools but which AI, what would a tool that is actually specific to teaching look like, and who should be in the room when it is designed?