In a recent conversation with educators, the meeting started where a lot of AI conversations start. What should students be allowed to use? Which tools should teachers try? Do we need a policy? What should count as cheating? How do we keep up? All of those questions matter, but after a while the conversation moved somewhere more important. What should students still be expected to think through for themselves? What kinds of assignments still make sense when a student can generate a polished response in seconds? What kind of learning do we want more of, not less of, as these tools become more capable?
At that point, we were not really talking about AI anymore. We were talking about learning. That distinction matters because schools can spend a lot of energy trying to build an AI plan. We can create guidance, evaluate tools, write policies, schedule training and define acceptable use. Those things have value, but if they are not connected to a deeper conversation about teaching and learning, they can easily become one more layer of activity without actually changing the student experience. The better starting point is not, “What is our AI plan?” It is, “What do we want learning to look like now?”
Start With Learning, Not the Tool
One of the easiest mistakes to make with AI is to focus almost entirely on what students should not do. Do not use it to write the essay. Do not use it to answer the questions. Do not use it to avoid the work. There are times when those guardrails are necessary, but if the conversation stays there, we miss the larger opportunity. The more useful question is what we want students doing more of. Do we want them asking better questions, explaining their thinking, making decisions, creating something original, defending a position, revising based on feedback, solving unfamiliar problems and connecting what they are learning to the world beyond school?
Those are the experiences worth designing for. Once we are clear about that, AI becomes easier to place. It can support the work, extend it, challenge it or make parts of it more accessible, but it should not become the purpose of the lesson. Technology should serve learning, not define it.
Remove the Wrong Kind of Friction
This also means recognizing that not all struggle is productive. Some difficulty comes from the thinking we actually want students to do. Other difficulty comes from unclear directions, lack of background knowledge, limited access or simply getting stuck before meaningful learning begins. Those are not the same thing.
AI can help remove some of the barriers that make learning harder without making it better. It can clarify directions, provide another example, offer feedback on an early attempt or help a student find a way into a task. Used carefully, those supports can help students reach the learning rather than replace it.
Protect the Thinking That Matters
The goal is not to make everything easier. The goal is to make sure the difficulty students experience is connected to the learning we actually value. That is where the conversation gets harder, because AI can remove barriers, but it can also remove the very thinking an assignment was designed to produce. A student can now generate a response that looks finished before they have wrestled with the idea. A polished paragraph is no longer reliable evidence that deep thinking happened.
That means some of our assignments need to change. If a student can successfully complete a task without understanding the content, the problem may not simply be AI. The task may need to be redesigned. That does not mean every assignment needs to become more complicated. It means we need to become clearer about where the thinking lives. Maybe students explain their reasoning before using AI. Maybe they compare their own thinking with an AI-generated response. Maybe they critique the response, improve it or defend why one answer is stronger than another. Maybe they use AI only after they have produced an initial idea of their own.
The question is not whether AI appears in the learning experience. The question is whether the student is still doing the work that matters.
Build a Learning Plan
This is why I keep coming back to the idea that schools do not need an AI plan first. They need a learning plan. An AI plan can tell people what they are allowed to use. A learning plan clarifies what we want students to become. That is a much more important conversation.
When principals lead this work, they do not need to arrive with every answer. They need to help their teams stay focused on a few enduring questions: What kind of thinking do we want students doing more of? What barriers can technology responsibly remove? What thinking, struggle and judgment must remain with the learner? Those questions are bigger than any single tool, platform or policy. They also give schools something stable to return to as the technology continues to change.
To me, that is what shaping the future really looks like: not chasing every new development, but helping schools decide thoughtfully what should change, what should remain and what students need most.
At your next staff or leadership meeting, take one common assignment or learning experience and ask three questions: What is the student really supposed to learn here? Which parts of this task could technology make more accessible or efficient? Which parts should students still own completely?
If those questions are clear, your AI decisions will become clearer too. You do not need an AI plan first. You need a learning plan.