What Your Planning Tool Forgets

Try this with any general AI tool you already use. Ask it for a lesson on fractions for your Year 5s. It'll give you something decent. Now come back in three weeks and ask for the next one. It has no idea who you are, what you taught, which explanation worked, or that four children in that class are still shaky on equivalence. You are a stranger again, and you have to re-explain your own classroom from scratch.

That re-explaining is the hidden cost of most AI in schools, and it's why the time saved never quite adds up to the time promised. Every prompt begins at zero. You become the memory — carrying context in your head and typing it in again, week after week, into a tool that never learns your room.

But teaching isn't a series of one-off requests. It's a long argument with itself. What you teach in November only makes sense because of October. The way you explain a concept the second time depends entirely on how it landed the first. A great department handover isn't a folder of resources; it's someone telling you what actually happened with this group last year.

So the question worth asking of any tool that plans with you is not "is the output good?" It's "does it remember?" Does it know what you covered, what you skipped, where the class wobbled and what you did about it? Because a tool with memory can do things a stateless one structurally cannot: notice that a topic keeps needing re-teaching, warn you that coverage has drifted from the curriculum, build Thursday's lesson knowing what happened on Tuesday.

This is the design decision underneath aime. It holds a teaching memory — your sequence, your classes, what was taught, what needed revisiting — so planning becomes continuous rather than a series of cold starts. You aren't briefing a stranger every Sunday. You're carrying on a conversation.

There's a legitimate worry here, and it deserves a straight answer: memory means data, and data about children demands care. The line we hold is that the memory is yours, it's about your teaching rather than a profile of a child, it stays within the boundaries your school sets, and it isn't used to train anyone's general model. If a tool can't tell you plainly what it keeps and why, that's the answer to whether you should let it keep anything.

The practical test is easy to run. Use a tool for a month, then ask it something that requires history — "which objectives has this class not securely met?" — and see whether it answers or asks you. That gap is the difference between a clever generator and something that actually knows your year.

Anyone can produce a good worksheet. Very little software knows what you taught last term, and that's the only thing that compounds.