AI can cut lesson planning time substantially, but honesty about the evidence matters more than a catchy promise. The strongest trial evidence so far points to a saving of about a quarter of planning time, not a half, and this article works from that honest number. It explains what the trial found, gives you a workflow that saves real time without lowering quality, and shows where the teacher must stay firmly in charge.
What the trial actually found
The stakes are easy to see: in England’s Working Lives of Teachers and Leaders survey, full-time teachers reported working 52.4 hours a week on average, and 48% of classroom teachers and middle leaders said lesson planning took too much of their time.
The strongest evidence comes from a randomised controlled trial run by the National Foundation for Educational Research and commissioned by the Education Endowment Foundation. One group of primary teachers planned with Aila, an AI lesson assistant that builds lesson plans and resources from a year group, subject, and topic. The other group was asked not to use Aila, although they were allowed to use other AI tools. What the trial found, at a glance, is set out in the table below, so you can compare the numbers without wading through them in the text.
| What was measured | What the trial found |
|---|---|
| Trial size | 464 primary school teachers across 108 schools |
| How the groups worked | One group planned with Aila, while the comparison group was asked not to use Aila, although they could use other AI tools |
| Length of the trial | Ten weeks |
| Planning time with Aila | Two and a half hours a week planning lessons and preparing resources |
| Planning time without Aila | Three hours and 19 minutes a week |
| Time saved | Around 49 minutes a week, about 24% less, with the Aila group spending 76% of the comparison group’s time |
| Quality of lesson resources | No difference, according to independent assessors |
| Confidence in the result | High security rating from the Education Endowment Foundation, meaning a high degree of confidence in the result |
| Use over the trial | 81% of the Aila group used it at least once, but use declined over the ten weeks, and it supported only a small share of lessons in a typical week |
| What teachers felt | Fewer than one in five teachers felt it had meaningfully changed their teaching or planning |
Read together, those numbers tell a reassuring story. The saving is real but modest, roughly a quarter of planning time, and it came with no drop in the quality of the lesson resources produced. That combination is what makes the result worth taking seriously: time returned, without lessons getting worse.
The honest caveats matter just as much as the headline. Teachers earlier in their careers, and those less confident in their subject knowledge, were more likely to benefit. Many teachers preferred adapting existing material to building lessons from scratch, which is what the tool is designed to do, and its output often needed editing before it was classroom ready. The lesson is not that AI planning failed. It is that AI planning saves the most time when it fits how teachers already work, which is usually adapting and improving, not starting from a blank page.
There is also an earlier trial worth knowing about. In a previous trial evaluated by the same organisations, teachers using a general chatbot for lesson planning spent 69% of the time spent by a comparison group asked to plan without AI tools. That trial also saw use decline over ten weeks. Taken together, the evidence supports a steady claim: AI planning help saves roughly a quarter to a third of planning time when it is used, and the saving depends on how well the tool matches the teacher’s real routine.
A planning workflow that saves real time
The workflow below is built around that evidence. It starts from your existing material, uses AI for the parts that are slow and mechanical, and keeps the teacher’s judgment at every step that affects children. Work through it in order, and resist the temptation to accept the first draft the tool produces.
Step 1: Start from your objective, not from the tool
Write one sentence stating what pupils should know or be able to do by the end of the lesson, and note the year group and the prior knowledge they already have. This sentence is the brief you give the tool, and it is also the test you apply to everything it returns. A tool given a vague brief produces a vague lesson. A tool given a precise objective produces something you can actually use, and the minute spent writing the sentence saves far more in editing later.
Step 2: Ask for parts, not a finished lesson
Instead of asking for a complete lesson plan, ask for the pieces that take you longest: a sequence of activities around your objective, a set of questions that move from recall to reasoning, a short explanation pitched at your year group, or a differentiated version of one task. Smaller requests are easier to check, easier to fix, and easier to reuse. They also match how teachers actually plan, by assembling and adapting pieces rather than accepting a whole lesson wholesale.
Step 3: Adapt it to your class before you use it
Read the output against your class, not against an ideal class. Change the examples to ones your pupils will recognise, adjust the reading level, remove anything that assumes knowledge they do not have, and add the misconceptions you know this topic usually produces. This is the step the trial suggests many teachers skipped or found awkward, and it is the step that turns a generic draft into a good lesson. Budget ten minutes for it, and treat it as planning, not as a chore.
Step 4: Build a bank you can reuse
Save the pieces that worked, with a note about what you changed and why. A bank of proven explanations, question sets, and activity sequences grows more valuable every term, because the second time you teach a topic you start from your own adapted material rather than from scratch. Over a year, that bank is where the larger time savings come from, and it is entirely yours, built from your judgment rather than borrowed from a tool.
Planning jobs, compared
The table below shows where AI help is strongest and where the teacher’s check matters most. Use it to decide what to hand over and what to hold on to.
| Planning job | Where AI helps most | The teacher check before use |
|---|---|---|
| Learning objectives | Drafting clear objectives from a topic and year group | Does it match the curriculum and what pupils already know |
| Lesson sequence | Suggesting an order of activities around one objective | Does the timing fit a real lesson, including settling and questions |
| Quiz and discussion questions | Generating varied questions from recall to reasoning | Are the answers correct, and do the hard questions actually test understanding |
| Differentiation | Producing simpler and harder versions of one task | Is the simpler version still the same learning, not a different and easier topic |
| Resources and handouts | Drafting explanations, examples, and practice items quickly | Is the reading level right, and are all examples appropriate for your class |

Where the teacher stays in charge
Three areas should never be handed over wholesale. The first is accuracy. AI tools can state wrong facts, wrong dates, and wrong methods with complete confidence, and a confident error in a handout teaches a whole class something false. Check every fact you have not verified yourself, especially in subjects outside your specialism, because the trial found that less confident teachers benefited most, which also means they are the ones least placed to spot an error.
The second is appropriateness. Only you know your class: reading levels, sensitivities, cultural context, and the pupils who need a different entry point. A generic activity that looks fine on screen can fall flat or cause real difficulty in a particular room. Read every output as if you were the pupil who will struggle most with it, and change it before the lesson, not during it.
The third is pupil data. Never enter pupils’ names, personal details, or identifiable work into a public AI tool. If you want help with feedback on a piece of work, remove every identifying detail first, or use a system your school has specifically approved for that purpose. The time saved by pasting a real script is never worth the privacy risk, and a general description of the common errors in a set of work gives you the same help without the exposure.
Making it stick across a term
The trial’s declining use is the most useful warning in the whole study. New tools get tried, then quietly dropped, usually because nobody built them into the weekly routine. The fix is small and practical. Pick one regular planning slot, such as the same afternoon each week, and use the workflow for one lesson in that slot. Share the pieces that worked with colleagues teaching the same year group, and review the bank together once a term to remove what did not work. A routine shared by a team survives far longer than a habit carried by one person.
The trial’s fading use is not unusual: in a RAND survey of US teachers in the 2023-2024 school year, only a quarter said they used AI tools for instructional planning or teaching.
It also helps to judge the workflow by the right measure. Do not ask whether AI made planning effortless, because it will not. Ask whether the same hour produces a better lesson with less strain, and whether the bank you are building makes next term easier than this one. On the trial evidence, a saving of around 49 minutes a week is a realistic expectation for a teacher who uses the tool regularly. Over a school year, that is time returned for marking, for preparation that genuinely needs a human, or simply for leaving school on time, and that is a result worth working for without exaggerating it.
Frequently Asked Questions
Can AI really halve my planning time?
The best trial evidence does not support halving. It supports a saving of about a quarter of planning time, around 49 minutes a week, for teachers who used an AI planning assistant regularly. A larger saving is possible for an individual teacher on a particular week, but a quarter is the honest figure to plan around, and it is still a meaningful amount of time across a term.
Will AI planning lower the quality of my lessons?
In the trial, independent assessors found no difference in the quality of resources produced with AI help compared with the comparison group. Quality held because teachers adapted and checked the output. The risk comes from using an AI draft without reading it, which is why the adaptation step in the workflow matters more than the tool itself.
What if I prefer adapting my existing lessons to starting from scratch?
You are in the majority, according to the trial, and the workflow in this article is built for you. Ask the tool for pieces that improve what you already have, such as better questions, a clearer explanation, or a differentiated version of one task, rather than asking it to build a whole lesson. That approach saves time and keeps your existing material at the centre.
Is it safe to put pupil work into an AI tool?
Do not enter pupils’ names, personal details, or identifiable work into a public AI tool. Describe the general pattern instead, such as the common errors in a set of answers, or use a system your school has approved for handling pupil data. The help you need rarely requires the identifying details.
How do I start if I have never used AI for planning?
Start with one small piece in one lesson: ask for a set of discussion questions on a topic you know well, check every answer yourself, and adapt the wording for your class. Once you have done that loop once, the rest of the workflow feels straightforward, and you will quickly learn which jobs the tool handles well for you.
Try it on one lesson this week
Pick the lesson you least enjoy planning, write your one-sentence objective, and run the workflow on that lesson alone. Check the output properly, adapt it for your class, and save the pieces that worked. If the hour feels lighter and the lesson holds up, you have a routine worth repeating, and a bank worth building. If your school wants a planning workflow designed around its own curriculum and tested in a small pilot before wider use, our Innovation Lab work starts with exactly that kind of practical pilot. For ready-made training sessions that build the same skills across a whole staff team, our AI Training sessions cover the workflow in this article step by step.


