Can AI Save Teachers Planning Time? A 2026 Study and a Four-Week School Pilot

International education research notes | 26 October 2026

Cover: a teacher reviewing digital and printed materials. All images in this article are AI-generated illustrations, not photographs of the research or actual school activities.

Preparing a lesson involves many small tasks that cannot simply be skipped: making instructions easier to read, designing questions at different levels, checking answers, and arranging materials so students can follow them. Evidence can help us judge where AI might share that work. For a school, the practical question is how much time remains saved after checking and revision.

A 2026 English study found time savings—with specific comparison conditions

In October 2026, the Education Endowment Foundation (EEF) released an independent evaluation of Aila by the National Foundation for Educational Research (NFER). The autumn 2025 trial involved teachers of Years 3–6 in England. Both groups could use other generative AI: it compared access to Aila with avoiding Aila. [1] EEF study overview

The main analysis covered 416 teachers in 108 schools. Adjusted weekly lesson and resource preparation averaged 2.51 hours in the Aila group versus 3.32 in the comparison group: about 49 minutes, or 24%, less. This does not measure reductions in total working hours or improvements in student learning. [2] Evaluation report, pp. 6–7

Resource quality was assessed after teachers could adapt materials; it met the study’s non-inferiority criterion. That is no endorsement of unchecked AI drafts. [1] Quality and implementation findings

Different percentages do not make a tool ranking

An earlier EEF randomised trial, reported in 2024, involved 259 teachers in 68 English secondary schools. For Year 7–8 science planning, ChatGPT with guidance saved an average 25.3 minutes weekly, or 31%, compared with a group avoiding generative AI. [3] EEF, 12 December 2024

Different year groups, subjects and comparison conditions mean that 31% and 24% cannot establish which tool is better. Schools should first specify the part of the workflow they want to improve, so they can later judge whether the change is worth keeping.

The OECD Digital Education Outlook 2026 likewise stresses clear pedagogical purposes and teacher involvement in educational uses of generative AI. Suitability still depends on a school’s own curriculum and students. [4] OECD, 19 January 2026

Start with a small task a teacher can evaluate

The following are Jing-Yeu International’s practical suggestions inspired by the research. They are neither the trials’ procedures nor a validated four-week intervention.

A first attempt might add three questions to existing material or rewrite one explanation at two reading levels. The teacher supplies accurate content and the learning goal, asks AI to help organise it, and checks the output item by item. A small task makes it easier to see which steps save time and which create extra revision.

For example, in a creative making lesson where students explain their designs for a canvas pouch, AI could help draft observation and reflection questions. Material suitability, production steps and equipment operation still require checking against local conditions and formal instructions. This is a proposed lesson idea, not a case tested in the research.

Caption: record generation, checking and revision together to assess the actual workload. AI-generated illustration.

A four-week pilot: turn impressions into records a team can discuss

Week 1 | Establish a starting point. Choose a small planning task within one subject and record the time and material version using the usual approach. Note existing tool habits too; there is no need to change the entire workflow just to make a comparison.

Week 2 | Change one step. Use AI for a comparable task and separately record prompting, waiting, checking, revision and formatting time. Use a school-approved tool and material that contains no identifiable student information.

Week 3 | Ask a colleague to try it. Invite a teacher in the same subject or year group to check three things: Is the content accurate? Can students understand it? Does the activity support the intended learning goal? Keep revised sentences and examples of errors for the next shared planning session.

Week 4 | Decide what to retain. Compare total time including checks, errors and revisions, and difficulties observed in class. If generation is quick but rewriting takes too long, narrow the task or change the approach. If one step consistently helps, turn it into an example colleagues can reuse.

A simple log is enough: date, lesson and task, approach used, total minutes, checking findings, final revisions and classroom observations. Four weeks of records can inform a local decision. Differences in topic difficulty and familiarity can affect the results, so a before-and-after change should not be treated as a causal AI effect.

A planning prompt to adapt

“Using only the lesson material I provide, design three questions for [year group/subject] with the learning goal of [specific goal]. Include comprehension, application and reflection, with suggested answers, supporting material and likely misconceptions. Do not add facts absent from the material. Where information is missing, list what the teacher needs to supply. Write in [teaching language]. The activity should fit within [number] minutes. Material: [paste content with personal information removed].”

This is our suggested prompt, not a standard prompt from the studies. Teachers should still verify answers and sources and try the activity themselves to check its difficulty and timing.

Bring the change back to the classroom

A practice worth keeping should make clear which task it supports, what checks it needs and which lessons it suits. Teacher workshops can start with a real piece of material from each participant: revise it once, exchange it for review, and record the steps they would choose to use again.

If your school is planning an AI teacher workshop or a creative making course, Jing-Yeu International welcomes a discussion about learning goals and lesson arrangements. Existing materials can also provide a starting point for practical tasks suited to your students. About Jing-Yeu International Explore creative materials and products

Sources and verification

This article is Jing-Yeu International’s original interpretation of primary sources, with complete Chinese and English versions. It is not an official translation issued by the research organisations. Figures and study conditions were checked on 26 October 2026. The practical suggestions, four-week pilot and sample prompt are original recommendations for this article.

[1] Education Endowment Foundation: Aila lesson-planning trial project page. Independent evaluator: NFER. The page records the report upload as 5 October 2026. Original project page

[2] Roy et al., NFER (October 2026). Lesson planning using AI lesson assistant, Aila – A Teacher Choices Trial. Key figures: printed pp. 6–7. Evaluation report PDF

[3] EEF (12 December 2024). Briefing on the ChatGPT-with-guidance lesson-planning trial. Original release

[4] OECD (19 January 2026). Digital Education Outlook 2026. DOI: 10.1787/062a7394-en. Official publication page