Does AI Reduce Teacher Workload? What School Leaders Can Measure

AI usually cuts the time a task takes without cutting the hours a teacher works. In a YouGov survey of 1,033 UK teachers for Bett, reported in August 2026, 51 percent of the teachers who use AI said it had eased their workload, yet only 35 percent said they were working fewer hours. The time came back, and something else filled it. In most schools nobody decided what.

That gap is the part a principal can do something about. You cannot make a model faster, but you can decide where the hour it frees will land, and you can find out whether it landed there.

Why do teachers say AI helps and still work the same hours?

Because a faster task and a shorter week are different things, and the surveys that report them ask different questions. The Bett survey, covered by Tes on August 27, 2026 and by Government Technology the next day, found that 80 percent of UK teachers now use AI and 52 percent use it weekly, up from 26 percent a year earlier. Among users, 51 percent said their workload had eased. Only 35 percent said they were working fewer hours. Government Technology reports that 55 percent work the same hours as before and 4 percent work more; it does not state the base for those two figures.

Government Technology's summary of the pattern is plain: "For most, the time saved on one job is swallowed by another."

Look at which jobs AI is doing. Bett reports that 76 percent of teachers who use AI rely on it to create lesson plans, worksheets and learning resources, and only 8 percent use it to mark or grade pupils' work. My reading, and it is an interpretation, not a finding: AI has landed first on the tasks where a teacher already had discretion over how much to do. A resource can always be polished further. A pile of marking has a bottom. The tasks that fill the evening are mostly the ones AI has not touched.

Is the "six weeks a year" figure wrong?

No, but it answers a narrower question than the headline suggests. Gallup and the Walton Family Foundation surveyed 2,232 US public K-12 teachers from March 18 to April 11, 2025, drawn from the RAND American Teacher Panel, and published the results on June 25, 2025. Teachers who use AI weekly estimated that it saves them 5.9 hours a week, which Gallup converts to about six weeks of the school year. Two details change how to read that.

First, only 32 percent of the teachers used AI weekly. Another 28 percent used it less often and 40 percent not at all. Averaged across all teachers, the saving Gallup reports is 2.3 hours a week in schools with an AI policy and 1.7 hours in schools without one, a difference the report does not claim is causal. Second, the 5.9 hours is a teacher's own estimate, net of any time AI added, and it is a statement about time on tasks. It is not a record of when anyone went home.

The two surveys do not contradict each other. One asks how long a job takes. The other asks whether the week got shorter. Both can be true in the same staff room on the same Friday.

Where does the saved time actually go?

It goes to one of three places, and only one of them is a lighter workload. The hour can go home with the teacher. It can go into a new task. Or it can go back into the same task, done more thoroughly than before.

Gallup's teachers describe all three. The report says they use the time for "providing more nuanced student feedback, creating individualized lessons, writing emails to parents and getting home to their families at a more reasonable time." One teacher told Gallup that AI "allows me to get my work done during my contracted time, so I am not having to stay late after work." That is the first destination. Better feedback is the third.

There is a fair objection here. If a teacher puts the freed hour into richer feedback, hasn't AI done something good? It has. It has improved the work. But a school that adopted AI to reduce workload, and told staff so, owes them the first destination and not only the third. When a leadership team reports "teachers say it saves hours" and stops there, it has credited AI with a workload benefit that the survey data suggests many teachers did not receive.

What is the Hours Ledger?

The Hours Ledger is three columns I ask a leadership team to fill in for any AI use that is meant to reduce workload: the minutes it takes out of a task, the minutes it puts back through checking and fixing, and where the net time went, whether home to the teacher, into a new task, or into more of the same task. A school that can fill in all three columns knows whether AI reduced workload. A school that can fill in only the first has a testimonial. I offer it as a suggested way of thinking, not a validated instrument.

Here is the ledger as a working sheet. Each teacher fills one row per task for two weeks.

A hypothetical example, to show how the columns interact. A history department drafts its Year 11 revision packs with AI. Each pack took 90 minutes. Drafting now takes 25. But the output needs its dates and quotations checked, and some of it corrected, which adds 30 minutes. The net saving is 35 minutes per pack. Then the head of department, seeing how quick drafting has become, asks for a second version of each pack for the intervention group. The 35 minutes has gone into a new task. Column one says 65 minutes saved. Column three says nobody went home earlier.

What is the mistake I see most often?

Across the leadership teams I work with, the mistake I see most often is reporting the first column and calling it the result. It is an honest mistake, because the first column is the one teachers volunteer. It is the satisfying one: "I used to take an hour and now it takes ten minutes."

As a school trustee, I read a workload claim the way a board reads any claim of improvement. Measured by whom? Over which weeks? Against what baseline? A story about one task answers none of those. Nor does it say what left the week to make room, because nothing did.

So decide the dividend before you collect it. When a department saves an hour, something must leave the schedule: a meeting that could have been a message, a data drop nobody acts on, a form that duplicates another. If you cannot name what is leaving, the saved hour will be taken by whoever asks for it first, and the person asking is usually a well-meaning colleague with a good idea.

How can a principal run the ledger in four weeks?

Pick one task, ask a small group to log it, and compare the log with a plain question about hours. Here is a plan that fits inside a half term.

1. Choose one high-volume task. Lesson resources, parent emails or report comments are common. Choose one where staff already use AI, so you are measuring practice and not a pilot.
2. Ask eight to ten teachers who use AI weekly to keep the ledger for two weeks. Include a few who do not use it, so you can see what the same task takes without it.
3. Add one question at the end of each week. "Compared with a typical week, did you work more, the same or fewer hours?" That is the Bett question, and it is the one that matters to a teacher's life.
4. Name the dividend in advance. Tell the group what you will take off their plate if the ledger shows net time saved, and then do it.
5. Share the results with staff in week four, including the rows where checking and fixing ate the saving. Treat the logs as estimates. Gallup's figures are estimates too.

Two supports make the measurement mean more. Guidance helps teachers know what to check, and it is thin on the ground: Bett found that 48 percent of teachers say their school has no written policy on staff use of AI, and where one exists, only 18 percent say they understand it well. Gallup found that 68 percent of US teachers had no school-provided AI training in 2024-25. If you are starting from there, what AI guidance teachers need, task by task and a short school AI policy are the first steps, and how to evaluate AI training for teachers shows how to check that practice changed.

What should you do with the numbers once you have them?

Report all three columns to staff and to your governors, and let the third column set the next decision. If the net time went home, protect that and extend it to the next task. If it went into a new task, ask whether the new task was worth the hour or whether it is the thing that should leave. If it went into more of the same, decide whether the quality gain is the outcome you wanted, and say so plainly instead of calling it workload relief.

A school that does this will know something most schools only believe. For the wider question of how you show a board that AI is making a difference, see how to prove AI is having an impact. To find out what staff think before you start the ledger, staff AI survey questions for school leaders is a good place to begin, and whether AI should write school reports tests the ledger on one of the heaviest tasks in the year.

If your leadership team is deciding where AI's saved time should go, this is the kind of work I support through AI strategy for schools. I also speak to school and system leaders and conference audiences on this theme, through my education keynotes.

Sources and further reading

• Turner, C. "Teachers' use of AI soars - but has little impact on working hours." Tes Magazine, August 27, 2026, reporting a YouGov survey of 1,033 UK teachers for Bett. Article.
• "Third of teachers say AI has not reduced their hours." Government Technology, August 28, 2026, reporting the same Bett and YouGov survey. Article.
• Bett UK. "The state of AI in education 2026," in partnership with Lenovo. Summary page; the full report sits behind a download form and was not read. Summary.
• Gallup and the Walton Family Foundation. "Teaching for Tomorrow: Unlocking Six Weeks a Year With AI." June 25, 2025, a survey of 2,232 US public K-12 teachers, March 18 to April 11, 2025, from the RAND American Teacher Panel. Report.

Dan Fitzpatrick is the founder of The AI Educator, a Forbes contributor, a school trustee and an international keynote speaker. More about Dan.