From designing exercises and lesson plans to checking answers and protecting privacy: a practical path to using AI responsibly in the classroom.
AI is useful in education when it helps teachers design, explain and review, not when it takes over responsibility for teaching and assessment. You can use it to draft a lesson plan or build exercises, but accuracy, fit for your students' age and the final decision still depend on the teacher's judgment.
In this guide we suggest a practical way to get started. The examples are Fekrano teaching patterns, not research findings or a promise of guaranteed time savings.
Start with one specific classroom problem
Instead of asking "what's the best tool?", pick a need you can observe: preparing exercises for two different levels, explaining a concept with several examples, or designing questions for the end of class. Choose the tool after you've set the learning goal.
Say the lesson is about fractions and part of the class still mixes up numerator and denominator. A good goal for a small experiment is three diagnostic exercises and one visual explanation. Producing a big pack of handouts and tests at the start makes errors harder to check.
Write a request you can check
A clear request names the topic, the audience level, the constraints and the output format. For example:
For a fifth-grade student, design three short exercises on comparing fractions. One exercise should reveal a common misconception. Write the answers and a step-by-step explanation separately. Don't use concepts outside this lesson.
This text is a starting point. If the output is too hard or vague, say exactly which part needs to change. Ask the model to explain its assumptions, but don't treat a long explanation as a sign the answer is correct.
Review the output like a draft before using it
- Solve the exercises yourself and check them against the textbook or another reliable source.
- Check the examples for language, age, culture and your students' experience.
- Fix ambiguous questions, options with two correct answers and overly complex explanations.
- Fit the text to your class's real needs; generic output isn't necessarily right for every student.
UNESCO's guidance emphasizes a human-centred approach, data protection and educational readiness for using generative AI. That means using it with review and human responsibility, not accepting output automatically. UNESCO guidance
Don't put student information into public tools
Building exercises rarely needs a student's name, phone number, photo, academic record or personal circumstances. Instead of uploading a real worksheet, write a made-up example with the same kind of mistake. Removing a name doesn't always make information anonymous; other details can still identify a person.
Before students use a service, check its age requirements and your school's policy. In class activities you can use output you've already reviewed, so a personal account isn't a condition for taking part.
Assess learning, not just the quality of the text
Fluent text can be produced without real understanding. Ask students to explain how they solved a problem, create a new example or find the mistake in an answer. Make clear at which stage using a tool is allowed and which part must be independent work.
For the suggested experiment, first run a short exercise without any tool. Then use the reviewed explanation, and finish with a different problem. Note how well students can explain their answers, what errors remain and how engaged they are. This small observation doesn't replace scientific research, but it helps you improve the next session.
Common questions from teachers
Can AI grade instead of the teacher?
It can help draft feedback or an assessment rubric. Handing the final decision to unreviewed output isn't appropriate, especially when the result affects a student's educational opportunities.
Do we need to know programming to start?
Not for the exercises in this article. Knowing the lesson goal, writing a clear request and being able to check the answer matter more. Which tool and features you need depends on the activity.
What's the first practical step?
Pick a topic from your next session, build three exercises, check the answers and use only the useful part in class. If you'd like to practice this in a workshop, see the AI for Teachers course or request a consultation.