AI lesson planning saved teachers time. They still didn't use it much.

NFER evaluated 464 Key Stage 2 teachers in 108 schools during the autumn term of 2025. Schools were randomly assigned to the trial groups. One group was encouraged to use Aila; the comparison group was asked not to use it. Both could use other generative AI tools.

That last detail matters when choosing software. This was not a comparison between teachers with AI and teachers without any AI. It tested adding encouragement to use this particular assistant in a setting where alternatives remained available.

NFER reports that the Aila group's lesson and resource preparation time was 76% of the comparison group's time. Its published summary gives around 2.5 hours a week versus 3.3 hours, with an estimated average saving of 48.6 minutes. Those are group results for the preparation work measured, not a promise to cut every teacher's total working week by a quarter.

The Education Endowment Foundation, which commissioned the evaluation, describes weekly teacher diaries supported by interviews and school visits. These are not stopwatch recordings of identical lessons. The randomised design makes the comparison valuable, while the way time was recorded still deserves to travel with the headline.

Aila helps teachers build lessons from a specified objective. Teachers select a year group, subject and topic, then refine the suggested objective and activities with the tool or offline.

That can be a useful start for an unfamiliar subject. NFER found that teachers newer to the profession or less confident in their subject knowledge were among those more likely to have lower workload in the Aila group. The evaluation also found mixed views: some teachers liked the curriculum alignment, while others preferred different tools or their existing planning methods.

The researchers suggest one possible mismatch. Aila is designed to help plan from scratch; experienced teachers may already have a lesson they want to adapt. Starting over can be unnecessary work even when the new plan is perfectly respectable.

This is a practical reason to ask what a teacher is stuck on before recommending an assistant. Needing a starting point for a new topic and needing to adjust tomorrow's familiar lesson are not interchangeable requests.

EEF reports that 81% of teachers in the Aila group tried the tool at least once, while both frequency and amount of use declined. Fewer than one in five respondents to a post-trial survey felt it had meaningfully influenced their teaching and planning.

Those findings should temper a rollout, but they do not erase the workload result. Nor does the workload result prove that declining use was healthy selectivity. People could have found a narrow use for it, preferred an alternative, or decided that adapting its output was too much trouble. The overall figures cannot tell a school which explanation fits its staff.

The quality finding also has a boundary: the lesson resources assessed were no worse in the Aila group. That is encouraging. It is not evidence here of improved pupil learning, and it does not mean a generated lesson needs no checking. NFER's report and EEF's announcement describe the same trial, not two independent confirmations.

For a teacher considering AI lesson planning, start with a forthcoming topic that would otherwise require substantial preparation. Keep a familiar lesson in its existing form unless there is a real reason to rebuild it. There is no benefit in manufacturing a blank page for the assistant to fill.

When judging the result, include the time spent changing the material before class. After teaching it, note anything that needed repair. A draft finished quickly but heavily rewritten later should not be recorded as an uncomplicated saving.

For school leaders, make the trial optional and keep the reporting light. Ask where the tool helped and where an existing resource was quicker. Do not require daily use as proof that staff are engaging: the published result itself came with limited and declining use.

Around 49 minutes less preparation in a week is worth taking seriously. Whether it becomes a less crowded evening depends on what happens to the rest of the workload. The sensible next step is to make one recurring preparation problem easier, then leave teachers room to decide when the assistant is worth opening.