Where Has the Work Gone?

A reflection on teaching, learning, and what counts as thinking in the age of AI.

Like many graduate students, I’ve been grappling with my use of AI in academic writing. Every assignment seems to raise the same question: How much AI is too much?

I’ve come to realize that’s the wrong question.

AI isn’t a shortcut for me. I don’t scrimp on effort or productive struggle. If anything, I spend more time wrestling with ideas than I did before. The final paper is still recognizably mine, yet it doesn’t feel like authorship in quite the same way it once did. AI isn’t making me think less; it’s forcing me to redefine what thinking is.

I also have to acknowledge something uncomfortable: the papers I produce with AI are probably better than the ones I would have written on my own. AI exposes me to perspectives I hadn’t considered, challenges weak arguments, and helps me articulate ideas more clearly. That raises legitimate questions about academic integrity, particularly when not every student has the same relationship with these tools.

But this isn’t another reflection on whether using AI is “cheating.” I don’t think that’s the most interesting question anymore.

The question I’m interested in is this: What counts as the work?

For a long time, I thought writing was the work. Read. Think. Draft. Edit. The effort was visible in crossed-out paragraphs, messy notes, and multiple revisions.

AI has disrupted that relationship.

One of its most deceptive qualities is that it produces work that looks finished. Designers have long understood that fidelity communicates progress. A rough sketch invites critique; a polished prototype suggests the important decisions have already been made. AI breaks that relationship. It can generate elegant prose in seconds, creating the illusion that an idea is far more developed than it really is.

The surface is polished long before the thinking has matured.

That realization has fundamentally changed how I write.

I’ve stopped treating AI’s first response as a prototype and started seeing it as a possibility—a provocation rather than a conclusion. Something to push against. Something to interrogate.

I question it. I argue with it. I ask it to defend its reasoning. I throw whole sections away because they don’t sound like me. I rewrite until the ideas reflect what I actually believe and understand, not simply what sounds plausible.

I’ve come to believe that ownership comes through iteration.

Not through writing the first sentence, but through deciding which ideas deserve to survive. Not through accepting suggestions, but through disagreeing with them.

In fact, I increasingly think your voice emerges by disagreeing with AI.

Perhaps that’s why working with AI feels surprisingly familiar. It resembles the design process far more than the writing process I learned in school. Designers know that the first concept is rarely the right one. Good work emerges through false starts, critique, experimentation, and countless small decisions. The creative act isn’t producing the first idea—it’s staying with it long enough to test it and make it your own.

The work hasn’t disappeared. It’s moved.

From production to judgement. From drafting to directing. From writing to iterating.

That shift has led me to a more unsettling question.

Am I developing a deeper understanding of my subject, or am I simply becoming exceptionally good at collaborating with a machine? Could I still synthesize journal articles without AI? Could I write a coherent argument entirely on my own?

I’m honestly not sure.

But perhaps we should question the purpose of those skills before we mourn their disappearance.

We teach students to synthesize literature, construct arguments, and write clearly—but those skills have never been the goal. They’re a means to an end. We value them because they help us build understanding and communicate it to others.

So if AI genuinely helps someone reach a deeper understanding of a complex topic, does it matter that they arrived there differently?

Of course, that’s only true if understanding actually happened.

The danger isn’t that AI writes for us. The danger is that it can create the appearance of understanding. It can produce convincing arguments without genuine comprehension. The challenge, then, isn’t deciding whether AI was involved. It’s determining whether the learner wrestled with the ideas, but that kind of wrestling doesn’t leave bruises. It isn’t always visible.

Perhaps that’s why AI feels so intertwined with our broader obsession with productivity. We live in a culture that celebrates efficiency and equates faster with better. AI promises more output, more quickly, with less effort. Universities often reward the same things: more assignments completed, more publications, more measurable achievements.

But learning has never been an efficient process. Neither has design.

Some of the most important moments happen while sitting with uncertainty. While abandoning a promising direction. While changing your mind. While allowing an idea to remain unfinished long enough to mature. Reflection, incubation, and productive struggle all look remarkably unproductive from the outside. Yet they are often where the real thinking happens.

AI doesn’t remove those moments.

Maybe that’s the work we need to protect, make visible and reward.

Not handwriting.

Not first drafts.

Not the performance of doing everything alone.

The effort to wrestle with ideas.

The willingness to change our minds.

The courage to reject elegant answers that don’t quite ring true.

If AI has made the work of thinking less visible, then perhaps the challenge for educators isn’t to police the use of AI, but to redesign learning so that the thinking becomes visible again.

Maybe that means asking students to show us their iterations rather than just their final submission. Maybe it means assessing judgment alongside writing, critique alongside composition, and reflection alongside results. Maybe it means creating learning experiences where the value lies not in producing an answer, but in defending it, questioning it, and changing it.

If we continue to assess only the polished artifact, AI will continue to blur the line between genuine understanding and its appearance. If, instead, we begin to value the invisible work—the false starts, the critique, the decisions, the moments of doubt—we may discover that AI hasn’t made education obsolete.

It has simply reminded us what learning was supposed to be about all along.

---

Disclaimer: This reflection was written with the help of my favourite thinking partner: the one who edits my grammar, does suspiciously efficient literature searches, but still needs me to tell it when it’s wrong.