Can we stop panicking for five minutes

Every article about AI in education falls into one of two categories. Either it's breathless enthusiasm ("AI will transform learning forever!") or it's abject terror ("Students will never write an essay again!"). Both are exhausting and neither is particularly helpful.

Here's the boring truth: AI is a tool. Like calculators. Like the internet. Like textbooks. It doesn't replace teachers any more than Wikipedia replaced libraries. But it can save teachers enormous amounts of time on the tedious parts of the job, time that can then go towards the parts that actually matter, like teaching.

I've spent the last few months talking to teachers who are quietly using AI tools in their classrooms. Not in a flashy, "innovation lab" way. In a practical, "I got my marking done by 6pm for the first time this term" way.

Here's what's actually working.

Generating lesson plans (and making them good)

The most time-consuming part of teaching isn't teaching. It's preparation. Creating lesson plans that are structured, differentiated, and aligned to the curriculum takes hours. AI can't do this perfectly, but it can produce a solid first draft in minutes.

The approach that works

ChatGPT and Claude both handle lesson plan generation well, but the prompt matters enormously. A vague prompt gives you a vague plan. A specific prompt gives you something genuinely useful.

Bad prompt: "Write a lesson plan about photosynthesis."

Good prompt:

That level of specificity produces lesson plans that are immediately useful. You'll still need to adjust them (AI doesn't know your classroom, your students' specific needs, or what you covered last week), but the structure is sound and the time savings are real. Teachers I've spoken to estimate they save 30-45 minutes per lesson plan.

Claude tends to produce slightly more creative activities. ChatGPT is better at sticking rigidly to the curriculum specifications. Both are worth trying to see which suits your style.

Creating differentiated worksheets

This is where AI saves the most time for the most teachers. Creating three versions of the same worksheet (foundation, core, extension) is the kind of tedious repetitive work that AI handles brilliantly.

The workflow Write (or generate) one worksheet at the middle ability level Ask the AI to create a simplified version with scaffolding (sentence starters, word banks, partially completed answers) Ask it to create an extension version with more challenging questions, open-ended prompts, and connections to broader concepts

ChatGPT is particularly good at this because you can share the original worksheet in a single conversation and generate all three versions without repeating context.

A real example

One maths teacher I spoke to creates differentiated homework sheets every week. Before AI, this took about two hours. Now she writes the core questions (about 20 minutes of work), feeds them into Claude with the prompt "create a foundation version with worked examples and sentence starters, and an extension version that includes problem-solving and real-world application questions," and gets usable drafts in about three minutes.

She still reviews and tweaks them. The foundation version sometimes pitches too low. The extension version occasionally includes content from the wrong year group. But the editing takes 20 minutes, not two hours.

Building quizzes with varied difficulty

Quick quizzes and retrieval practice are one of the most evidence-based teaching strategies we have. The problem is that writing good quiz questions takes time, especially if you want a range of question types and difficulty levels.

The prompt that works

Both ChatGPT and Claude produce good quizzes from this prompt. The multiple choice questions are reliable. The short answer questions are fine. The application and extended response questions sometimes need adjustment because AI tends to make them either too broad or too specific. But again, editing is faster than creating from scratch.

Using Canva for visual quizzes

Canva has a quiz/presentation template that works well for classroom display. You can use AI to generate the questions, then pop them into a Canva presentation with timers and visual elements. It looks more polished than a plain Word document and takes about the same time once you've got a template set up.

NotebookLM for study materials

NotebookLM from Google is quietly becoming one of the most useful tools for education. Upload a textbook chapter, a set of notes, or a research paper, and NotebookLM will create summaries, generate study questions, explain difficult concepts, and even create audio overviews that students can listen to.

How teachers are using it

The most common use I've seen: uploading textbook chapters and having NotebookLM generate study guides. The AI reads the content and creates a condensed version with key terms highlighted, comprehension questions embedded, and summary boxes for each section.

One history teacher uploads primary sources into NotebookLM and has students ask the AI questions about the documents as part of a guided analysis exercise. The AI provides accurate answers based on the uploaded source (not from the internet), which means students are engaging with the actual material rather than just Googling.

The audio overview feature is surprisingly popular. NotebookLM generates a podcast-style discussion of uploaded material, which some students prefer over reading. It's particularly useful for revision: students can listen to summaries of topics while commuting or doing chores.

The limits

NotebookLM only works with content you upload. It won't pull from the internet or generate content from nothing. This is actually a feature for education, because it means students are engaging with curated materials rather than random internet content.

Addressing the cheating concern (briefly, because it's complicated)

Yes, students use AI to write essays. This is a real problem and pretending otherwise is silly. But it's also not a new problem. Students have always found ways to avoid doing work they don't want to do. AI just made it easier and more convincing.

The practical responses that seem to work:

Change the assessment. More in-class writing, more oral presentations, more project-based work that requires demonstrating process, not just product. If the only evidence of learning is a take-home essay, you're going to have a bad time.

Teach AI literacy. Show students how to use AI as a tool rather than a replacement. "Use ChatGPT to brainstorm ideas, then write the essay yourself" is a legitimate workflow that develops both AI literacy and writing skills.

Focus on the learning, not the submission. If a student submits an AI-written essay and then can't discuss it intelligently in class, the essay grade matters less than the obvious gap in understanding. Design assessments that test understanding through multiple channels, not just written submissions.

I'm not going to pretend there are easy answers here. There aren't. But the teachers I've spoken to who've adapted their assessment methods are coping much better than the ones trying to detect AI use after the fact.

The practical toolkit

For a teacher who wants to start using AI without overhauling everything: ChatGPT or Claude (pick one) for lesson plan drafts, worksheet generation, and quiz creation Canva for making those materials look presentable NotebookLM for creating study materials from existing content Grammarly (optional) for cleaning up reports and emails

Total monthly cost: potentially zero, if you stick to free tiers. ChatGPT and Claude free tiers are sufficient for most teaching tasks. Canva has a free tier for educators. NotebookLM is free.

The real impact

The teachers using AI most effectively aren't doing anything revolutionary. They're doing the same things they've always done, just faster. They're not replacing their professional judgment with AI output. They're using AI to handle the repetitive, time-consuming grunt work so they can spend more time on the things that actually require a human being in the room.

And that, frankly, is exactly how AI should work in education. Not as a replacement for teaching, but as an assistant that handles the paperwork while teachers do what they actually trained for: helping young people learn.

If you're a teacher reading this and feeling overwhelmed by AI, take a breath. You don't need to change everything. Start with one thing. Use AI to draft next week's lesson plan. See if it saves you time. If it does, try the next thing. If it doesn't, that's fine too. The tools aren't going anywhere, and there's no rush to adopt all of them at once.