The 3 Questions Every Teacher Should Ask Before Using AI in the Classroom

There is a moment that happens before almost every AI decision in a classroom, and it rarely gets talked about. A new tool shows up, and a student asks if they can use it. A district rolls out a policy that leaves the actual judgment call to you, and in that moment, most teachers reach for the same unspoken question: “Is this okay?”

That question is understandable, and it is also the wrong one. "Is this okay" turns every AI decision into a rules problem, something to check against a policy document that was written before this particular assignment, this particular student, this particular Tuesday morning existed. The policy cannot keep up. It was never going to.

What actually holds up is not a longer policy. It is a smaller, sturdier set of questions that travel with you into any classroom situation, because they are not about the tool. They are about the learning. That is the CALM AI Framework: three questions, asked before AI enters a lesson, a task, or a decision.

Question 1: What is the learning goal?

Before deciding whether AI belongs anywhere near an assignment, name precisely what the assignment is protecting.

• Is it a skill: the ability to construct an argument, to solve for x, or to read a primary source closely?

• Is it a piece of knowledge: dates, vocabulary, or the structure of a cell?

• Or is it a habit of thinking: persistence through a hard problem, or the discipline of drafting badly before drafting well?

This sounds simple, and it is the step teachers skip most often, usually because the pressure of the moment asks for a fast answer, but a vague goal produces a vague decision. "I want them to write well," does not tell you whether AI helps or harms. "I want them to practice building a thesis from evidence, unassisted," does. The precision is the work. Once the goal is precise, the next two questions answer themselves far more easily.

Question 2: What belongs to AI, and what belongs to you?

This is the question that separates support from replacement, and the distinction matters more than any policy line about "permitted tools."

Support means the AI scaffolds the thinking:

• it organizes notes the student already generated;

• it offers a second set of eyes on a draft the student already wrote;

• it handles a repetitive step so the meaningful step gets more attention.

Replace means the AI does the thinking the assignment exists to build, and the student submits the appearance of learning instead of the learning itself.

The same tool can do either, depending entirely on when and how it is used. A student who asks AI to brainstorm counterarguments before drafting an essay is doing something different than a student who asks AI to write the essay and edits the transitions. Same tool. Different question answered. The teacher's job is not to memorize every tool's capabilities. It is to hold the line between scaffolding and bypassing, assignment by assignment, and to say so plainly to students before they have to guess.

Question 3: How do we keep learning at the center?

This last question is the check on the first two. Whatever gets decided, whichever way the tool gets used, the decision should protect the learning, not just make the workflow easier or the grading faster. Ease is not the wrong goal entirely; invisible labor is real, and relief matters, but ease is the wrong first question. When it becomes the only question, the assignment quietly stops measuring what it was built to measure, and everyone, teacher included, loses the thread of what is actually being taught.

Keeping learning at the center sometimes means the harder answer: no, not on this one, because this is exactly the struggle the assignment is designed to build. Other times it means real relief: yes, here, because this step was never where the learning lived anyway. The question does not produce the same answer every time. It produces a consistent way of deciding.

What this looks like on an ordinary Tuesday

None of this requires a new policy, a faculty meeting, or a semester of retraining. It requires a teacher willing to ask three questions before the tool comes out, instead of after something already feels off. What precisely is the learning goal? What belongs to AI, and what belongs to me? How does this decision keep learning at the center.?

Some answers will be imperfect. That is allowed. A decision made from three clear questions beats a decision made from panic, and it beats a decision borrowed wholesale from whoever posted the loudest opinion online this week. The goal was never to have AI figured out completely. It was to know how to think it through, one classroom decision at a time.

This is the same framework behind the one-page CALM AI Framework. It’s designed to sit on a desk or get taped up where a teacher can glance at it before Tuesday morning's decision arrives. It is available as a free download, and it is the same thinking that runs through every course and conversation in the AI Faculty Break Room, a free community for educators doing this work without the hype and without doing it alone.

If this is the question that has been missing from your own classroom decisions, start with the framework. Download the CALM AI Framework here, and if you would like company while you put it into practice, the AI Faculty Break Room is free, and it is waiting.