Mr. Okafor teaches Year 8 history in a school with 32 students per class and one shared laptop trolley. Last September, he did not roll out a new AI platform. He changed one 10-minute routine. Students now paste a short source into a chatbot, ask for a simpler explanation, then hunt for what the AI left out. That small habit led to better questions, fewer blank stares, and a staff meeting where other teachers wanted to try it.
That is what practical AI in schools looks like. It is less about buying the shiniest product and more about building a few repeatable steps. If you are still working out what the term even means, a realistic look at classrooms powered by artificial intelligence is a good starting point. This guide takes you from audit to student-led projects in seven concrete steps.
Step 1: Audit What You Already Have
The best AI schools start with an audit, not a purchase order. Before your school spends a dime, find out how much AI is already in the building. Most staff use tools with AI features without calling them AI. Google Workspace suggests replies. Microsoft 365 can transcribe meetings. Canva generates slide layouts. Reading apps adjust difficulty based on student answers.
Run a 30-minute audit in your next staff meeting. Ask every teacher to name one digital tool they used in the last five lessons. Then mark which ones adapt, generate, or grade automatically. A primary school I know found that its reading programme already used AI to set book levels. No new purchase was needed. They just trained teachers to use the existing reports.
Four Questions for the Audit
- Does the tool adapt content based on student responses?
- Does it generate text, images, or questions?
- Does it send student work to a third-party server?
- Which single tool could you trial for two weeks without extra budget?
Step 2: Write a One-Page AI Policy with Students in the Room
Policy documents often die in a shared drive. A one-page agreement that students help write actually gets used. At Riverbend High, the draft policy says AI is allowed for brainstorming, outlining, and practice questions. It is not allowed for final essays. Students must add a short note when they use AI: ‘I used a chatbot to generate practice quizzes for this unit.’
The policy also covers what parents will ask about. Cheating. Screen time. Data privacy. If you need plain-language answers for families, what is real, what is hype, and what parents need to know gives you a calm script. Share it at back-to-school night before rumours start.
Three Questions Every Policy Should Answer
- What is allowed? Name specific tasks, not vague principles.
- What must be disclosed? A one-line note is enough.
- What is banned? Uploading student IEPs, generating images of real people, using AI to write final submissions.
Step 3: Start with One Routine, Not a Whole Curriculum
Teachers do not need another 40-hour training course. They need one routine they can run tomorrow. Mr. Okafor’s routine is called ‘explain, then improve’. He pastes a three-paragraph historical source into an AI tool and asks for a Year 8 reading level. Students read both versions side by side. Then they list what the AI simplified, what it got wrong, and what it erased. The final step is rewriting one sentence themselves.
That routine works in science, geography, and English. A biology teacher uses it to compare a textbook paragraph with an AI summary of photosynthesis. A music teacher uses it to generate practice questions about chord progressions, then asks students to correct the AI’s mistakes.
For staff who feel behind, Duke’s AI for Everyone course is a non-technical way to build confidence. It takes a few hours, not a semester.
Step 4: Build a Feedback Loop That Saves Teacher Time
Grading is where AI can help or hurt. Use it for first-pass feedback on low-stakes work, never for final grades without review. A science teacher I spoke to generates five multiple-choice questions from each lab report using a chatbot. She edits two of them, deletes one, and loads the rest into a quick quiz. Students get immediate feedback while she writes one deep comment per report instead of twenty shallow ones.
Students can run the same loop at home. Using Meta’s AI as a study partner shows how to ask for quizzes, explanations, and follow-up questions without letting the tool do the thinking. The key is teaching students to prompt for practice, not answers.
Where AI Feedback Goes Wrong
- It rewards formulaic answers and penalises creative risk.
- It misses cultural context and local examples.
- It can invent sources or dates with complete confidence.
- It cannot know a student’s mood, effort, or recent progress.
Step 5: Teach AI Literacy Inside the Subject
AI literacy is not a separate lesson. It is a habit you build into normal work. In history, ask an AI tool to describe a local event. It will likely invent a date or a detail. Students then verify with two real sources. That is the lesson. In maths, ask AI to solve a word problem and show its work. Students check each step and find the one where it hallucinated.
For a ready-made skills path, Intel’s free AI learning hub covers prompt basics, model limits, and ethical use. You can assign one module as homework and discuss it the next day.
A Five-Minute AI Literacy Check for Any Subject
- Ask AI a factual question you already know the answer to.
- Find one error or oversimplification.
- Rewrite the prompt to get a better answer.
- Name one thing the AI still cannot do.
Step 6: Measure What Changed and What Did Not
After a term, look at three numbers. How many minutes of admin did teachers save per week? How many students can explain one AI limitation? How many AI-related academic integrity referrals did you have? A maths department I know saved about three hours per teacher per week on marking practice questions. Essay scores stayed flat. That is a useful result. It means the tool improved workload, not learning, and the school can decide whether that trade is worth it.
Run a small pilot with one class and one comparison class. Do not claim AI raised attainment unless you can show it.
Step 7: Let Students Lead the Next Iteration
The final step is handing over the controls. Ask a group of students to run a lunchtime workshop for teachers on one AI tool they actually use. They will show you shortcuts, risks, and workarounds that no professional development session covers. At one school, students built a prompt library for revision, then wrote a one-page guide for Year 9. Teachers adopted it because it came from people who had to sit the exams.
After that, the cycle repeats. Audit again. Update the policy. Try a new routine. Drop the tool that wastes time. AI schools are not a finished product you install. They are a set of habits a school keeps tuning, with students and teachers in the same room.