The Two Jobs You Keep

Almost everything can be handed to the tool. Two things cannot.

Hand a student AI and it will do most of the work. Draft the essay, summarize the reading, explain the hard part, tidy the sentence. Almost all of that is fine to give away. Two things are not. One is the moment where a person decides what they actually think. The other is the habit of checking whether an answer is true. Those two are where judgment is built, and they are the two a tool is quickest to take. Everything that follows is about keeping them.

The deciding

The deciding is the moment, somewhere inside any real piece of thinking, where a person settles on what they actually believe. What is true here. What is good. What they are willing to put their name to. It passes quickly, which is exactly why it is easy to lose. AI is unusually good at filling it, offering a finished view before one has formed. When that happens, nothing looks wrong. A reasonable position arrives, it gets used, and the student simply never has the ordinary experience of arriving at it themselves. That experience, repeated across years, is most of what judgment turns out to be.

The checking

The checking is quieter still. An answer that sounds certain is not the same as an answer that is right, and the distance between the two is the space a careful reader learns to live in slowly. It is unglamorous work, and it is the first thing to go when the output already looks polished. To keep it, you must treat what the tool says as a claim rather than a conclusion. To lose it you simply stop noticing there was a difference.

Why the amount of time was always the wrong measure

Hold onto those two and the rest can be handed over freely. Give them away and it hardly matters how sparingly the tool was used elsewhere. A student can lean on AI constantly and keep both, or lean on it for the single decision that mattered and quietly hand them over. That is why the amount of time was always the wrong measure. The question is not how much of the tool a student uses. It is which parts of the thinking stay with them.

One question that tells you

There is a plain way to tell, and it is closer to a question than a rule. After using it, could you defend the result without the tool in the room? When the answer is yes, the tool did what tools are for. If it is no, something was handed over that was probably worth keeping. No one can score that question except the user. Noticing the difference may be the most anyone can honestly ask of themselves here.

What this asks of a parent or a teacher

None of this requires expertise, which is the part worth saying to any parent or teacher doing this without a map. You are being asked to guide students through something you did not grow up with, often while being told the only choices are to ban the tools or to teach a course on them. Neither is required. You are not competing with the tool. You are keeping present the two important steps it is happy to make invisible.

For a teacher, that can be as small as asking what a student thought before they typed their prompt, and then walking the path rather than jumping straight to the destination. For a parent it can be a simple question: what was your student curious about today, and what did they do about it? What both are doing is treating the student as someone capable of deciding and thinking, rather than someone being taught at. Students tend to rise to whichever version of themselves they are handed.

That is most of it. The tools will keep getting better at the tasks, and they should. What is worth protecting is narrower and quieter than the noise around it: the moment of deciding, and the habit of checking. They were never tasks to be finished. They sit close to the person doing the thinking, which is the whole reason it matters that the person keeps them.