3 AI Tools I Actually Use Every Week (and One I Ditched)

The honest version

I have tried a lot of AI tools over the last 18 months. Some have genuinely made my week lighter. Some looked shiny in a screenshot and then disappointed me by Wednesday. And one I quietly uninstalled and did not tell anyone about until now.

This is not a "10 amazing tools you MUST try!!" listicle. It is the actual handful I open every week, and what they are genuinely for. Grab a brew.

The tools that earn their place

A quick thing first: I have left out brand names on purpose. Tools come and go; the job they do is what matters. Whatever AI you have got open right now can do all three of these jobs beautifully if you ask it well.

1. A planning assistant

The job: turning a clear learning objective into a full lesson skeleton I can then adapt.

I do not ask AI to write a whole lesson blindly. I give it a thought-out prompt: the learning objective, the year group, the time, and the bits I want. For example, a starter, worked example, guided practice, independent task and plenary. Then I let it draft the bones.

Then the teacher layer goes back on top. I tweak the example so it speaks to my class. I switch the stimulus for one I know lands. I add the questioning I have been polishing for a decade.

The draft saves me 40 minutes. The teaching is still mine.

2. A differentiation helper

The job: generating support, core and stretch versions of the same task, then checking the stretch actually stretches.

This is the one I think gets misused most. It is so easy to ask for a "harder version" and end up with the same task plus three extra questions. That is not differentiation. That is volume.

So I always add one line to the prompt: "The stretch version should deepen thinking, not add quantity. Show me how." Suddenly the stretch has higher-order verbs in it, a metacognitive prompt, a justification step, the kind of thing my top-end pupils actually need. Little prompt tweak, big lift in quality.

3. A parent communication drafter

The job: getting the tone right when the stakes feel high.

Firm. Warm. Specific. Professional. All at once. On a Friday. After period 5. We have all stared at that blank email for twenty minutes.

I use AI to draft the awkward ones: a behaviour concern, a missed assessment, a sensitive pastoral nudge. Not because I cannot write them, but because a first draft on the page is so much easier to edit into shape than a blank screen is to write into.

Key trick: tell it the tone you want, the relationship history, and what you do not want to imply. Never enter specific data. The output gets noticeably more human.

Where AI still needs a teacher

This is the bit I want to be really clear on, because the discourse around AI in schools often misses it.

I do not trust any tool to know my class: the names, the friendships, the pupil who has just lost a grandparent, the one who is brilliant but will not speak in pairs.

I do not trust any tool to know my school context: the curriculum we have built, the language we use, the routines that make a lesson run.

I do not trust any tool to know the misconceptions my pupils carry into the room: the half-learned thing from last term that is quietly tripping them up.

That is the teacher layer. It stays human. It always will.

AI gets you a better starting point. You are still the lesson.

The one I ditched

The tool I quietly stopped using was an AI marking and feedback engine: the kind that promised to read pupils' work, mark it, and write personalised feedback comments for me. Tap a button. Done. Sleep well.

The comments came back polished. They were also disconnected from anything I had actually been teaching. They missed the misconception we had worked on all term. They praised things I had never set as success criteria. And they did not sound like me. Pupils notice that quicker than we think.

Here is the line I have landed on: feedback, marking, knowing your class - that is the most teacher bit of teaching. It is the bit AI should not be doing for you. It can help you draft a faster comment bank. It can sense-check a tricky paragraph. It can save you the blank-page panic. But the eyes-on-the-work step? That stays with us.

I would rather mark thirty books with a tired pen than outsource the bit where I actually see my class.

The pattern

Notice what all three keeper-tools have in common: they do not try to replace the teaching. They quietly remove the admin scaffolding around it. The first draft. The differentiation grunt-work. The blank-page email.

That is the difference between AI that helps and AI that hinders. The good stuff knows it is the understudy, not the lead.

The honest summary: AI did not come to replace you. It came to give you your Sunday back.