Holi, Soy Anita.

Subscribe free newsletter and get the workflow templates →

Let me tell you about a version of a story I’ve heard from more teachers than I can count, and have lived by myself more than once. 

A teacher buys an AI tool. It’s a grading assistant, a lesson generator, something that promises to hand back your week. Week 1, it’s magic. The tools spit out a worksheet, a rubric, a set of comments, in seconds. Then week 2 arrives, and something’s off. The AI-generated feedback doesn’t quite match how this teacher actually grades. The lesson plan needs restructuring because it doesn’t fit how their students actually learn. The “5-minutes task” has quietly turned into 45 minutes of rewriting, fact-checking, and translating the AI’s generic output into something that sounds like an actual human who knows these actual students. 

By week 3, the teacher is doing everything by hand again, except now there’s also a subscription renewal reminder sitting in their inbox as a small monthly reminder of the experiment that didn’t work. 

Here’s the thing nobody tells this teacher: the tool probably wasn’t the problem. The missing workflow was. 

Using a tool and integrating a tool are not the same purchase 

This is worth saying plainly, because the 2 gets treated like synonyms constantly, and they are not. Using technology means the tools are present in your teaching. Integrating it means the tool is actually built into how you plan, teach, and evaluate with a defined role, a defined handoff, and clear boundaries around what it does not do for you. 

Research on classroom technology backs this up in a very  unglamorous way: what predicts whether a tool actually helps isn’t how often it gets used. It’s the quality of how it’s built into the teaching itself. A tool opened 5 times a week with no real workflow behind it will consistently underperform opened twice a week that’s actually slotted into a designed process. 

Which brings us back to the 5 layers from the first post in this series. Most tool adoption is only ever tough layer 2 which is the tool itself. What’s missing is layer 1 ( the design, deciding what this tool’s job is, before you open it) and layer 3 (the coaching, deciding how you personally get fluent enough with it that it stops needing constant supervision.) 

What integration would have actually looked like 

Go back to our teacher and the grading assistant. Here’s the same tool, with 15 minutes of layer-1 design added before it ever touched a single student’s work: 

• Decide its exact job, narrowly. 

Not “help with grading.” Something specific: “draft first-pass comments on grammar only, leave content feedback to me.” A narrow job is a job tool that can actually do well.

• Decide what you’ll never hand it. 

Anything requiring knowledge of this specific student’s history, or judgement calls about effort versus ability, stays with the teacher. That’s not the tool being bad at its job. That’s the boundary that keeps it useful. 

• Build the two-minute check into the plan, on purpose. 

Not “hope it’s right and find out later.” A specific, expected step; skim every AI-drafted comment before it goes out. That 15 secs per student was always going to happen, the only choice is whether it’s designed in from the start or discovered, annoyed, in week 2. 

None of that required a better tool. Same subscription, same AI, completely different outcome because the 15 minutes of design work happened before the tool did, instead of never happening at all. 

The part where you’re not actually to blame. 

If this is you right now, sitting on a tool you paid for and quietly resent, I want to say clearly: this isn’t a personal failure of tech-savviness. Almost nobody teaches this distinction, because almost every AI tool’s marketing page sells you layer 2 and never mentions that layer 1 and 3  exist at all. You were sold “just use this,” and “just use this” was always missing half the instructions. 

The fix usually isn’t a new tool. It’s going back to the one you already have, and giving it 15 minutes of a job description before you open it again. 

I’m building out the actual templates for this, how I scope a tool’s job before I let it near my own workflow over in the newsletter with The Shine Lab. If this explained a tool you’ve been quietly resenting, that’s exactly where the next layer of this lives. 

Subscribe and get the workflow templates →

— Anita