The marking and lesson prep week
For teachers and school owners. Where your hours actually go, why AI helps least with the thing everyone expects it to fix, and how to run the whole thing in one sitting a week.
Written August 2026. Prices and free tier limits change often, so check them yourself before committing to anything. Nothing here requires a paid subscription.
The situation
A teacher with a class of fifty is not short of skill. She is short of hours. In most schools I have worked with, the week outside the classroom goes on four things: writing lesson notes in the format the head expects, setting and marking tests, writing report card comments at the end of term, and answering parents. Term ends and the comments alone can mean fifty short paragraphs written after nine at night, each one supposed to say something true and particular about a child.
School owners feel it differently. Good teachers leave because the administrative load never lets up, and the ones who stay are too tired to plan properly. So the question is not really about technology. It is about giving hours back to people who need them.
Does AI actually help here?
Partly, and not where most people expect.
The assumption is that AI will mark the scripts. It will not, and you should not let it. Marking student work means judging a particular child against what you taught and what you know they can do. A model does not know your syllabus, your marking scheme or your class, it is confidently wrong often enough to matter, and getting a grade wrong is not a small error to a family. There is also the point below about never uploading children's work in the first place.
Where it genuinely earns its place is the writing that surrounds the teaching. First drafts of lesson notes. Turning one lesson into three versions for a mixed ability class. Building a question bank with a marking scheme you then correct. Turning your own rough notes about a child into a readable report card comment. In every one of those, you supply the judgement and the model supplies the typing. That is the whole trick, and it is why this works on a small budget: you are not buying intelligence, you are buying speed on the parts that were always mechanical.
The rule you cannot break
Never put a child's name, photograph, school record or scanned work into an AI tool. Use initials or "Student A". Keep anything identifying on paper or on your own device. These services are not covered by your school's duty of care, you cannot promise a parent where that data went, and the convenience is never worth it. Everything in this guide is designed to work without a single identifying detail leaving the room.
The setup
Assume one device, one sitting, and data you are paying for by the bundle. The aim is a single session of about ninety minutes a week, with everything else done offline.
- Collect offline during the week. Keep a notebook or a notes app on your phone. For each child, jot two or three specifics as they happen: "struggles to start, fine once moving", "reads well, rushes the last question". Two lines a child across a term is enough, and it costs no data because you are writing, not searching.
- Write your prompts before you go online. Type them out in your notes app first. This is the step most people skip and it is where the data goes: opening a chat and thinking aloud in ten separate messages costs far more than pasting one considered instruction.
- Do the online part in one batch. Paste, generate, copy the output straight into a document on your device, move to the next task. Do not browse. Do not regenerate hoping for something better.
- Close the connection and edit offline. The editing is the real work and it needs no internet at all.
On tools: use a free tier and build nothing that collapses when a trial ends. The paid tiers are priced in dollars, which for most Nigerian teachers is the actual barrier rather than the amount. If a guide anywhere tells you to subscribe to five services, close it.
The prompts
These are starting points, not magic. The instruction that matters most in each is the one forbidding invention, because a model asked to write about a child it knows nothing about will happily make something up.
Report card comments
You are helping a secondary school teacher in Nigeria write end of term report card comments. I will give you short rough notes about each student, identified only by initials. Write one comment per student, three to four sentences, in British English, warm but honest, addressed to the parent. Use only what my notes contain. Do not invent achievements, grades, incidents or personal details. If my notes are too thin for a full comment, say so and tell me what to add rather than filling the gap yourself. Each comment must end with one specific, doable suggestion for next term. Vary the sentence openings so the comments do not read as a template. Here are my notes:
Lesson notes
Write a lesson note for a [subject] class of [age or year group] in a Nigerian secondary school. Topic: [topic]. Duration: [minutes]. Learning objectives, which I have set and which you must not change: [objectives]. Structure it as: objectives, materials needed, introduction, main activity, class exercise, evaluation questions, assignment. Assume a class of [number] with limited equipment and no reliable projector. Keep the language plain enough to read aloud. Do not add topics beyond the objectives I gave you. If any part of the topic needs a fact you are not sure of, mark it with CHECK rather than guessing.
One lesson, three ability levels
Here is a lesson note I have already written. Produce three versions of the class exercise only: one for students who are behind and need the concept broken into smaller steps, one for the middle of the class, and one that stretches the strongest students without introducing new topics. Keep the same objectives and the same vocabulary. Do not change the main teaching part of the lesson. Lesson note follows:
Question bank with a marking scheme
Create [number] test questions on [topic] for [year group], mapped to these objectives: [objectives]. Mix recall, application and one question that requires explanation. For every question also give the marking scheme, showing where each mark is awarded. Number the questions. Do not include a question that depends on knowledge outside the objectives listed. Flag with CHECK any question where the correct answer depends on a specific syllabus detail I should verify.
What goes wrong
- Every comment sounds the same. This is a symptom of thin notes, not a fault in the tool. Two genuine specifics per child fixes it immediately.
- It invents things. Even with the instruction above, it will occasionally add a prize or an incident that never happened. You read every line before it goes to a parent. That is not optional, and it is why this saves you typing rather than responsibility.
- The curriculum drifts. The model does not reliably know the Nigerian syllabus. Always supply the objectives yourself rather than asking it what should be taught, and treat anything marked CHECK as unverified.
- It writes beautifully and says nothing. Fluent, warm, empty. If a comment could be about any child in the school, it has failed and you rewrite it.
- The data disappears faster than expected. Almost always because of chatting back and forth rather than pasting one prepared instruction.
Prove it to yourself before you trust it
Do not take my word for the time saved, and do not let anyone sell you a figure either. Measure it. For two weeks, write down the hours you spend on lesson notes, question setting and report comments. Then run the method above for two weeks and write the same numbers down. If it has not given you back real hours, drop it without regret. If it has, you now have evidence rather than a claim, which is also the thing to show a head teacher or a parent who asks why the school is using this at all.
If you run a school
Two things are worth doing before this spreads through your staffroom on its own, which it will. Write down the rule about children's names and give it to every teacher, because the first time a member of staff pastes a class list into a chat window is a problem you want to have prevented. And ask your teachers to keep their own before and after hours for a term. If it genuinely returns time, that is a retention argument and a case for training. If it does not, you have saved yourself from buying something that was never going to work.