If the Final Product Can No Longer Prove Learning, What Should?

FIELD NOTE #006 · 📍Destination: Instructional Leadership · Prioritize

For a long time, schools could make a fairly reasonable assumption. A student turned in the essay. Completed the project. Solved the problems. Built the presentation. Wrote the response. And while none of those things ever gave us a perfect picture of what a student knew, the finished product usually gave us some evidence of the thinking that happened along the way.

AI has complicated that.

📍Current Reality

A polished product no longer tells us as much as it used to. A student can submit writing that sounds stronger than what they can explain. A presentation can look thoughtful without the student having made the decisions behind it. A solution can be correct without showing whether the student understood why it was correct.

And that creates a much bigger school-leadership problem than cheating. The real question is: if the final product can no longer prove learning, what should?

That question is already surfacing in education policy conversations. A September commentary about Ohio's new school AI policies argued that acceptable-use rules aren't enough — schools also need a "proof-of-learning" standard, a way to distinguish between AI helping students learn and AI simply helping students produce.

I think that distinction matters. Because we can spend a lot of energy trying to answer "did the student use AI?" when the more important question may be: can the student demonstrate what they understand?

🔄Missed the Turn

If the essay alone is no longer enough, we may need to look at more of the learning process — not because every student needs to prove they're innocent, but because the work itself may no longer contain enough evidence for us to make a strong judgment about learning. That could mean drafts and revisions, short oral explanations, conferences, demonstrations, applying the learning to a new problem, explaining why a source was chosen, showing how an answer changed, identifying which AI suggestion was rejected and why, or completing part of the work in a setting where the teacher can observe the thinking.

That may create better evidence of learning. It also creates more work. If we tell teachers to review the drafts, check the revision history, conference with the student, listen to the oral explanation, compare the process to the final product, ask follow-up questions, document AI use, and verify whether the student can transfer the learning — that's a real cost, and it matters just as much as the assessment design itself.

More trustworthy evidence usually requires somebody's time. And schools have to decide whose.

💡New Information

This isn't theoretical — educators are already experimenting with these approaches. At the University of Saskatchewan, one instructor redesigned a writing course around multiple checkpoints, source verification, opportunities for students to explain their thinking orally, and feedback throughout the writing process, after finding that some students were relying on AI-generated summaries, fabricated sources, and AI-written essays. The University of Virginia has compiled guidance around oral exams and oral assessment as one way educators can get closer to what students actually understand in an AI-rich environment. And emerging assessment frameworks are increasingly shifting weight away from the polished final product and toward reasoning, process, and students' ability to explain or defend their thinking.

One recent framework for AI-resilient assessment estimated that oral defenses for 100 students could require roughly 25 to 35 hours of total faculty time once preparation, scheduling, review, and moderation are included. That's higher education, not K–12, so I wouldn't treat that number as a direct estimate for a classroom teacher. But the implementation lesson absolutely transfers: more trustworthy evidence usually requires somebody's time.

That's why I keep coming back to a question school leaders should ask whenever we talk about AI efficiency: when AI saves time for one person, where does the work go? Sometimes it disappears — that's real efficiency. Sometimes it shifts. And sometimes it multiplies. Good implementation requires us to know the difference.

We also have to be careful not to confuse "proof" with surveillance. If schools become obsessed with catching AI use, we could build assessment systems around suspicion — prompt logs, screenshots, version histories, AI detectors, declarations, constant proof that students did not cheat. One recent analysis of AI-resilient assessment makes this point clearly: process evidence should be collected when it helps educators understand learning, not simply as forensic paperwork. A student explaining how their thinking changed can deepen learning. A student producing six screenshots only because adults want evidence of innocence may not.

The goal shouldn't be "prove you did not use AI." It should be "show me what you understand, how you got there, and what you can do with it." That's a very different learning environment.

🧭Next Right Move

This is where leadership matters. It's easy to tell teachers "you need to redesign your assessments for AI." It's much harder to answer: what are we going to stop doing so they have time to do that well? Which assessments truly need stronger verification? Where would an oral explanation add value — and where would it simply create more work? Can one checkpoint replace three pieces of paperwork? Could we redesign the assignment itself rather than adding another layer after it's complete? Can students provide evidence of thinking in ways that also deepen learning, instead of turning into forensic documentation? What technology can responsibly reduce the administrative part of this work without taking over the instructional judgment? And which expectations are we willing to remove because they no longer make sense in an AI-enabled classroom?

Those are leadership questions. Not teacher-compliance questions.

AI didn't create the problem of overvaluing polished products. It exposed it. Schools have always had students who could produce something without deeply understanding it. We've always had assignments where completion looked like mastery. We've always had situations where a grade told us less than we thought it did. AI simply made the gap harder to ignore — and that may be an opportunity, if we respond well: more opportunities for students to explain, more emphasis on reasoning, more application, more feedback during learning instead of only after it, more attention to whether students can transfer what they know. But only if the system underneath it is designed to carry the work.

So this is the leadership question I'd bring back to my own team: don't start with "how do we stop students from using AI?" Start with "what evidence would convince us that the student actually learned?" Then ask how much time collecting that evidence will require, who will carry that work, what can be removed or simplified to make room for it, which parts need human judgment, which parts technology can appropriately support, and — maybe most honestly — are we creating better learning, or simply more documentation? (I built out a more complete framework for sorting exactly that kind of judgment call in AI-Ready Does Not Mean AI Everywhere, if it's useful here too.)

Because if the finished product can no longer prove learning, we absolutely need better evidence. But we can't keep adding new expectations to classrooms as though teacher capacity is unlimited. That's not implementation. That's accumulation. (This is the same distinction I made in What Did It Take From People to Produce That Number? — a stronger system and a more exhausted staff can both produce a "better" result, and only one of them holds.)

If we want assessment to become stronger in the age of AI, we have to redesign both sides of the equation: how students demonstrate learning, and how schools create the capacity to see it.

Real Support. Real Systems. Real Results.

If your school is heading into these questions and you want a clearer picture of where the real gaps are, take the free School Leadership Reality Check — a clear, honest look at where to start.