Here is an exercise I run with students, and it works better than any rule I've tried to hand them.
A paper is due Friday. The question is how the sun affects our sleep. The process looks like this:
- Read what the syllabus requires
- Gather the research
- Create your outline
- Write the first draft
- Edit and finalize
- Submit
AI could do every one of those. So: where does it belong?
I don't give them the answer, because the argument is the point. But here's roughly where I land, and where I'm still unsure.
Step 1: Read the requirements
Yours. Obviously.
Though it is worth noticing that "summarize this syllabus for me" is a real temptation, and that the person who reads the actual requirements catches the thing the summary dropped. Small stakes here. The habit matters more than this instance.
Step 2: Gather the research
Give it away, mostly.
This is where AI is genuinely strong and where the friction was mostly waste. Finding sources, pulling out relevant sections, orienting yourself in an unfamiliar literature. There is no version of you that's a better thinker because you spent four hours on database search syntax.
One condition: you have to actually read what it finds. AI-assisted gathering followed by human reading is a real improvement. AI-assisted gathering followed by AI-assisted summarizing followed by writing about things you never read is how you end up confidently wrong in public.
Step 3: Build the outline
This is where I genuinely go back and forth, and I tell students that.
The case for handing it over: an outline is structure, and structure is a solved problem. There are only so many ways to organize an argument.
The case against, which I find more persuasive on most days: deciding what your argument is and what order it goes in is the thinking. The outline is where you find out whether you have anything to say. Getting one handed to you means you skipped the part where you would have discovered that you did not.
My honest position is that it depends on whether you already know what you think. If you do, let it help you organize. If you do not, doing it yourself is how you find out.
Step 4: Write the first draft
Keep this one.
This is the rep. Everything upstream was preparation and everything downstream is refinement. This is the step where a person turns a pile of material into a position, and it's the step where the growth actually happens.
It is also the least pleasant step, which is exactly why the temptation is strongest and why the trade is worst.
Step 5: Edit and finalize
Share it.
Having a second reader is not new and was never cheating. Every good writer has had someone tell them the third paragraph is muddled. AI is an available, patient, slightly generic version of that.
Use it to find problems. Fix them yourself. There is a real difference between "point out where this is unclear" and "rewrite this to be clearer," and the first one leaves you better at writing while the second leaves you with a better paragraph.
Step 6: Submit
Yours, and this is the one people treat as trivial.
Submitting is a claim. You are saying this is mine and I stand behind it. If steps two through five were done by something else, that claim is false regardless of what any policy says about it.
The pattern underneath
Notice what the answers have in common.
The steps worth handing over are the ones where the work is retrieval, formatting, or mechanics. The steps worth keeping are the ones where the work is judgment.
That isn't a rule about school. It is the same test for anyone in any job. Which parts of this are logistics, and which parts are the reason someone hired a person?
Automate the first. Guard the second. And be suspicious of your own reasoning when you find yourself concluding that the unpleasant part was logistics all along.
Why I don't just give students a rule
Because a rule teaches compliance and this exercise teaches discernment, and they are going to need the second one for the rest of their lives in situations I can't anticipate.
Also because when they argue it out, they are better at it than I expect. They already sense which steps are the ones that make them. They just haven't been asked.
Try it on your own work this week. Take something you do regularly, break it into six steps, and mark which ones you would hand over. The disagreements you have with yourself are the useful part.