There’s a student I think about when I talk about differentiation. Let’s call her Wei. She spoke English as her third language, was sitting in an AP World History course, and was producing work that was technically correct but somehow flat — like she was translating thoughts rather than thinking in the language of the essay.
She wasn’t struggling because she lacked intelligence or effort. She was struggling because the cognitive load of writing in a second (or third) language was eating up the bandwidth she needed for historical analysis. The thinking and the writing were competing for the same resources.
This is the differentiation problem that AI actually helps with — not the simple accommodations, but the genuinely hard ones that teachers with thirty students can’t solve through individual attention alone.
What Differentiation Actually Requires
The honest version of differentiation is that most of us aren’t doing it well, and not because we’re bad teachers. Genuine differentiation requires knowing where each student is, what they specifically need, and having the time to design for it. In a class of thirty, that’s not realistic at scale.
What most teachers do instead is modify at the margins — bigger font, extra time, simplified vocabulary on handouts. These help. But the student like Wei, who needs a genuine scaffold for the cognitive demands of writing analytically in her non-primary language, isn’t really being reached.
AI changes the math here in a specific, practical way: it makes one-to-one scaffolding possible at classroom scale.
How It Works in Practice
Here’s what I’ve found most useful. I’ll share an essay prompt with a student like Wei and then also share it with an AI model, along with context: This student is writing in English as a third language, working on AP-level historical analysis. She understands the content well. Help her build an outline that separates the thinking from the writing — one where she can work out her argument in simpler language first, before translating it into formal essay structure.
What comes back is a scaffold, not a crutch. It helps her get her thinking onto the page in whatever form it takes, and then gives her a framework for shaping that into the required register. The intellectual work is still hers. The cognitive bottleneck — translating between a language of thought and a language of performance — gets a little wider.
For students with different needs, the prompts are different. A student with dyslexia who understands the material but struggles with organization might get a different kind of scaffold. A student who is academically advanced and needs more challenge gets a different kind of push. The AI is a tool for generating individualized supports, not a replacement for the teacher who knows which student needs what.
The Part AI Can’t Do
I want to be careful here, because there’s a real limit to this.
AI can generate a scaffold based on the description you give it. It cannot observe a student in the moment, notice that she’s checking out during discussion, and realize that she actually understood the lesson better than her silence suggested. It cannot know that a particular student had a hard week at home, or that he responds better to questions than to instruction. It cannot replace the accumulated judgment of a teacher who has been watching a human being think and grow all year.
What AI enables is more of the individualized support work that used to be genuinely impossible at scale. It doesn’t know your students. You do. But it can help you serve them better once you’ve decided what they need.
A Shift Worth Making
The deeper shift here is about what we think differentiation is for. If we think it’s about lowering the bar for some students, AI-assisted scaffolding looks like enabling shortcuts. But if we think differentiation is about removing structural barriers that prevent students from demonstrating what they actually know — which is what it’s actually for — then AI becomes a legitimate tool for equity.
Wei didn’t need a simpler question. She needed a temporary scaffold that let her get her actual thinking out without fighting the language at the same time. That’s the difference. And it’s one that teachers have always known, even when they lacked the tools to act on it.
This post draws on ideas from The AI Doesn’t Know Your Students, Chapter 10: “The Student in Front of You.” The book is available now.
