With AI, students for the first time now have access to tools that can perform substantial portions of the intellectual work schools ask them to do.

In 2026, this is not news. Educators have been wrestling with this reality for at least the last three school years.

Students can generate an essay, summarize a reading, solve a problem, draft code, explain a concept, create an image, prepare a presentation, and increasingly help with the planning and reasoning that precede those products.

While all this creates extraordinary possibilities for education, it also creates a problem schools won’t solve by banning AI use. The problem to which I’m referring is embodied in this question:

As AI becomes increasingly capable of doing cognitive work for students, what intellectual work do we still need students to own?

I suspect this will become one of the defining educational questions of the next decade.

A human-centered school is not an anti-AI school

I do not think the appropriate response is to protect students from artificial intelligence. Our young people today will live and work in a world saturated with it. Our graduates should become exceptionally capable of using AI to extend what they can know, create, investigate, and accomplish.

And teachers should use it where it makes their work better. They should experiment. Students should learn how to collaborate with increasingly capable machines.

But there is an important distinction between using AI to extend human capability and allowing it to replace the human capacities we were trying to develop in students in the first place.

Education is about becoming powerful.

If a student uses AI to identify possible weaknesses in an argument and then decides whether the criticism is valid, AI may be strengthening the student’s thinking.

If the student asks AI to decide what the argument should be, construct it, defend it, and then polish the prose, something very different has happened.

Writing teaches people to organize ideas, wrestle with ambiguity, discover what they think, make distinctions, anticipate objections, and communicate with precision. We ask them to read difficult texts because sustained attention, interpretation, and encounters with ideas unlike their own can change the way they understand the world. It would be a shame if those human capacities are not developed in our students.

When we are at our best as educators, we ask students to struggle because that struggle is part of learning. AI creates an interesting educational challenge because while it is potentially so empowering, it is extraordinarily good at removing struggle.

Schools will need to get good at providing learning experiences that help students to use AI appropriately while maintaining the struggle that leads to powerful learning.

Student agency is now more important than ever

The more powerful AI becomes, the more persuaded I am that our central educational goal should be developing students’ agency.

Agency means knowing what I want, understanding why I want it, examining whether it is worthy, and retaining ownership of the judgments that guide my actions.

Human ownership of thinking is going to be especially important in an AI-shaped world. As people are increasingly surrounded by systems willing to recommend, predict, draft, decide, personalize, optimize, and answer, the temptation will be to allow the technology to do the thinking. There is a possibility that it will become normal to surrender small pieces of judgment because doing so is easy, useful, and usually works.

For we educators, the profound educational question is:

How do we prepare students to use extraordinarily capable tools without gradually surrendering ownership of their own thinking?

This is a much more important question than the one I hear often when speaking to teachers: “should students be allowed to use ChatGPT?”

Much thinking in an AI-shaped world must remain human

Schools need to think through, and develop, better answers to questions about which cognitive work can be productively delegated and which should remain with the learners. A good guiding question for those conversations might beL

What thinking must remain deliberately human because the act of thinking is itself formative?

Schools will find that the answer will not be identical across ages, disciplines, or learning goals. There won’t be universal rules that fit across a K-12 school.

For this reason, I expect attempts to create one AI policy will likely fail. Instead, schools need to foster conversations among teachers about what human capacities should be built in each learning experience, and how AI can help or should be excluded.

Three questions schools should be asking

I have found three questions especially useful in conversations with educators and school leaders:

What should students become exceptionally capable of doing with AI?

What should they remain capable of doing independently?

What relationships, experiences, and forms of productive struggle become more valuable as AI becomes more capable?

These three questions lead quickly into curriculum, assessment, academic integrity, faculty practice, strategy, and institutional governance.

When considered and discussed earnestly, those questions lead us to think more clearly about the human work of school. They help us identify the slippery slope of allowing AI to automate, accelerate, and produce where students should be doing the thinking and production work.

Likewise, those questions can lead us into conversations of what teachers should stop doing in their practice, and what approaches they should be using instead.

Schools must remain human institutions

Schools’ value has always depended partly on things that are inefficient by design: conversation, mentorship, disagreement, practice, revision, relationships, reflection, community, and the slow development of judgment. Yes, AI can make much of that easier. But is doesn’t mean human effort is less valuable.

As AI becomes increasingly more capable, schools need to become more intentional about creating experiences in which students exercise judgment, encounter difficulty, take intellectual responsibility, listen to other people, defend ideas, revise beliefs, and discover what they themselves think.

To be clear, AI should change education, and it already has. The challenge to us as educators and school leaders is to make sure that as we redesign education we remain clear about what we are trying to develop in the human students entrusted to us. We do not yet know exactly what AI will be capable of three, five, or ten years from now, but we now know enough to continue to ask this important question:

What do we want our students to become?

---

Stay in the conversation

I’ll be writing regularly about AI in education, student agency, academic integrity, curriculum, assessment, school leadership, and governance as schools work through these questions. Subscribe to receive new essays when they are published.