The Process and the Residue

It goes like this. I'll be chatting with a colleague — could be over a random WhatsApp chat, could be in a fly-past walk-by (you know, both places to go, but you manage to drop a few sentences as you pass like ships in the night…), could be one of those slightly-too-honest after-school conversations — and at some point they'll say something along the lines of: "The thing is, it looks like learning is happening. They're using AI, they're engaged, the work looks great. But… how do I know if they’re actually learning anything?"

How do I know if they’re learning? I keep hearing that phrase, or versions of it, and it's starting to haunt me.

And then, separately, I was doing some pupil voice work with a group of students, and one of them said something that properly stopped me in my tracks. I'm paraphrasing, but it was basically: "Well, if you smashed it on a take-home essay where you used AI but then you can't pass the exam or repeat it in class, you've clearly not learned anything, have you?"

Folks, the students get it.

They know. They know there's a gap between what the AI helped them produce and what they can actually do on their own. And here's what struck me: they shouldn't have to wait until an exam to find that out. Neither should we.

Because we spend so much time talking about the process of using AI well — how to prompt, how to think critically about output, how to iterate — that I think we've been neglecting a much more fundamental question:

What actually remains once the laptop is shut?

The Process and the Residue

I've started thinking about this as the difference between the Process and the Residue. The Process is everything that happens on the screen — the prompting, the back-and-forth, the critical evaluation of AI output. It's important. I've literally written a whole book about it (move along, nothing to see here, just a casual plug).

But the Residue is what stays in the student's head after the screen goes dark. It's the bit that actually matters for learning. And right now, I think we're spending a lot of time perfecting the Process and not nearly enough time checking for the Residue.

Here's the thing: a student can do everything "right" with AI — brilliant prompts, critical evaluation, thoughtful iteration — and still walk away having learned absolutely nothing.

The Process can look fantastic whilst the Residue is basically zero.

There's a specific version of this that we're seeing a lot more It goes like this: a student works with AI and produces something that uses sophisticated academic language, complex sentence structures, and domain-specific terminology. It’s also, importantly, level/age-appropriate, containing the information you’d expect them to know. In other words, we’re not talking about cheating here. This is definitely the student’s work, albeit AI-enhanced.

Basically, it reads like they understand.

But what's actually happened is they've absorbed the AI's vocabulary without internalising the concepts behind it.

So how do we — and more importantly, they — start to recognise what the Residue actually is, and what they might need to work on more?

Screens Down, Brains On

Here's your new golden rule, and it's stupidly simple: before any residue check happens, the screens go down. Laptops shut, tablets flipped, phones away. Every time. No looking at what you might already have written or prompted. True learning requires recall, and recall only happens when the source material is invisible. So the screens go down. Every time. Non-negotiable. Make it a ritual, make it a routine, make it the thing your students groan about but secretly know is coming. "Miss, do we have to close the laptops?"

Yes. Yes you do.

And then what? Well, here are three low-tech, high-impact techniques that you can use to check whether anything has actually stuck. I'm calling them the Sticky Checks, partly because they all happen on sticky notes and I think that’s cute, and partly because they’re all about what has stuck in the brain post-AI.

Before we look at them, there’s one more thing: this doesn’t have to happen directly after the AI session. It’s not compulsory to have them as Exit Tickets. In fact, it might even be better NOT to do that. Give them a bit of space. See what really sticks. 

And one more bonus: mini whiteboards also work really well here. Unfortunately, I couldn’t think of a cute name for MWB-related residue checks, but I’m sure you can come up with one...

1. The Iconic Summary

Imagine this, but even worse drawing, and even better learning...

Instead of asking students to write a paragraph summarising what they've learned (which, let's be honest, often just becomes a memory test of the AI's phrasing), ask them to draw three icons and write exactly five words that capture the core idea.

Hand out a post-it. Laptops shut. Three icons, five words. Go.

This works because of dual coding — you're forcing the brain to convert linguistic information into visual information, creating two separate mental pathways to the same concept. The student has to translate, not just repeat. And here's the smell test: if a student can't draw an icon for it, they don't understand it. You genuinely cannot draw a symbol for "macroeconomic stability" unless you actually know what it looks like in practice. Try it. I'll wait.

Yes it’s hard. Yes you’ll have students complain that they can’t draw. It’s okay, you can join in the fun. Use a visualiser to show the worst pictures. I promise, the worse the drawing, the stickier it’ll be. 

2. The Curriculum Bridge

(Side-note, Nano Banana really let me down on 'different handwriting on each post-it note' and also, in general, the concept of writing... and post-it notes.....)

Ask the students to write one sentence connecting the AI's output to something specific from a previous lesson. Not "something we've learned before" — something specific. Yesterday's starter. The case study from Tuesday. That thing we argued about last week.

"Write down one way what the AI just told you contradicts — or supports — what we talked about yesterday."

This is schema building in action. AI provides information in a vacuum. It has no idea what happened in your classroom yesterday, what examples you used, what that one kid said that made everyone laugh and accidentally illustrated the concept perfectly. Only your students know that. Only you know that. This is where the AI fails and the teacher wins, every single time.

When a student can hook new information onto existing knowledge — their existing knowledge, from their classroom experience — that's when encoding actually happens. That's when the Residue starts to form. The AI can give them information, but it can't give them the connections that make it stick.

3. The Telephone Test

Fun fact I just learned - if you take a screenshot of only a section of the AI generated image (like I did for the first two pictures in this article), no AI credit shows up on LinkedIn. It only shows on the fully uploaded image. Hmmm...

In this one the student has to repackage the AI's conclusion — usually sophisticated, often jargon-heavy — into a single sentence for a specific person. A seven-year-old. Their grandparent. A historical figure. Someone who definitely does not speak in AI.

"If you had to explain the AI's main point to a Year 2 student on the playground, what ten words would you use?"

This is basically the Feynman Technique wearing a school uniform. To simplify a complex idea without losing its essence, the brain has to perform genuine synthesis. The student has to strip away the AI's vocabulary and find the actual core of the concept. If they can do it, they understand it. If they can't — if they're stuck trying to rephrase the AI's language rather than expressing the idea in their own words — then the Residue isn't there yet.

I love this one because you can get some really silly examples that you’re going to want to hold onto for future reference. Students explaining quantum physics to an imaginary seven-year-old produces some genuinely brilliant (and occasionally unhinged) results.

Keep it Going: the Residue Board

These post-its don't just stay on desks. They go up on a board — an Exit Ticket Board, or stuck to the classroom door as students leave. Over time, you can start to build a visible map of what's actually sticking.

The board becomes a conversation piece. Students start looking at each other's post-its. They argue about whether someone's icon actually captures the concept. They steal each other's five-word summaries and improve on them. The Residue starts generating its own Residue. Which, if you think about it, is basically just... learning. 

Yep, once again, we’re succeeding in tricking our students to learn better. I’d call that a win.

---

Look, none of this is revolutionary. Mini-whiteboards and post-it notes aren't exactly cutting-edge pedagogy. But that's sort of the point. The most powerful check on whether AI has actually supported learning isn't another digital tool — it's the oldest trick in the book: close the laptop, pick up a pen, and prove you actually know something.

We've been so focused on making the AI process better — and we should keep doing that — but we need to be equally obsessed with what's left when the process is over.

The Residue is the learning. Everything else is just... screen time.

So next time your students finish an AI-assisted task and it all looks beautiful and polished and impressive, try this: shut the laptops, hand out the post-its, and ask them to show you what's actually in their heads.

You’ve got the Process down. Now let’s check out the Residue.