A student hands in an essay, a lab report, or a page of solved problems. It is the thing we asked for, so naturally we read it closely. We look at the answer, the reasoning that made it onto the page, the organization, the mistakes, and the quality of the finished work. For most of my teaching career, that final product had to carry nearly the whole weight of telling me what a student understood.
Even before AI, that was never quite enough. A polished paper might reflect deep understanding, a heroic round of revisions, extensive help, or a student who had learned how to sound convincing without being very sure what the words meant. A messy paper might come from a student with an excellent idea who ran out of time, got tangled in the writing, or had not yet figured out how to make the thinking visible. The final answer mattered, but it never told the whole story.
Now that AI can produce a highly polished answer in seconds, the gap between a finished product and the learning behind it has become impossible to ignore. That does not mean the final product is useless. It means teachers need a wider window.
Look at the path, not just the destination
Think about what happens between the moment a student first sees a problem and the moment the work is submitted. The student makes an attempt, gets stuck, tries an approach, receives feedback, changes something, explains a choice, and perhaps tries again. Those moments are not clutter surrounding the real work. They are where much of the learning actually happens.
A useful process record might help a teacher see the student's first attempt, the question or hint the student opened, what changed afterward, and whether the student could explain the revision. Over time, it might also show whether the student needed the same amount of help again or began to handle a similar task more independently. None of those observations is dramatic by itself, but together they can give a teacher a better place to begin.
The key phrase is a better place to begin. A process record should lead to a question, a conference, a teaching decision, or another opportunity for the student to try. It should not turn into a machine-generated verdict about the student.
Evidence is not a verdict
Suppose a student pasted a large block of text into a draft. That is an observable fact. It is not, by itself, proof of cheating, proof that AI wrote the work, or proof that the student learned nothing. The text might have come from notes, an approved accommodation, a shared planning document, or something the student wrote elsewhere. The record gives the teacher something worth asking about. It does not supply the answer.
The same is true when a student opens several hints, pauses for a long time, or makes a quick revision after receiving feedback. Those actions may help us understand the path through an assignment, especially when we know the student and the classroom context. They cannot tell us the student's intent, emotional state, ability, authorship, or overall level of understanding.
This distinction matters because the technology can make a thin slice of information look much more authoritative than it really is. Put a number next to something, add a colored warning, or call it a risk score, and suddenly an observation begins to feel like a conclusion. Teachers and students deserve better than that.
Ask the student about the work
The most useful questions are often ordinary ones. Why did you choose this evidence? What were you trying to do in this paragraph? What changed after you received feedback? Which part are you still unsure about? If you tried a similar problem tomorrow, where would you begin?
These questions do more than help a teacher investigate the work. They ask the student to become part of the explanation. A student who can describe a choice, notice a weakness, and identify a useful next step is doing important thinking, even if the finished product is still rough.
That is the kind of wider window I want Coach to help provide. Not a system that claims to know what happened inside a student's mind, and certainly not an automated detective. I want it to preserve useful moments from the learning process so that teachers can ask better questions and students can better understand their own work.
The final answer still matters. We should read it carefully, respond to it, and expect students to produce work they can stand behind. We just should not mistake the final page for the entire story of how learning happened.