There’s a lot that goes into a teacher’s job beyond just teaching. There are all the permission slips, replying to parent messages, logging behaviour notes, and prepping tomorrow’s lessons, while also somehow remembering that this coming Friday is non-uniform day (again).
So, when I find myself reading about companies working on agentic AI systems, the first question that comes to mind is whether the latest tech could give teachers a few minutes a day back without making school feel like a robot factory.
We don’t need schools to become futuristic. We just need classrooms to feel calm and supportive, and for teachers to be able to breathe.
Workload Problem Getting Louder
Ask any teacher, and you’ll hear some version of the same story: the job spills past the bell. Planning, marking, admin, data entry, safeguarding notes, meetings, emails, more emails, and the kind of mental load that follows you into the supermarket queue.
If you want a stark snapshot of this, the UK Department for Education’s Teacher Workload Survey is worth a read.
In that survey, teachers reported working 49.5 hours in the reference week on average, with full-time teachers and middle leaders reporting 52.9 hours.
Source: https://assets.publishing.service.gov.uk/media/5e12fcb7e5274a0f9e82e4fd/teacher_workload_survey_2019_main_report_amended.pdf
Zooming out for a moment, it’s helpful to remember that a lot of a teacher’s work week is not about standing at the front and teaching. The OECD has a short, visual summary that breaks down teaching and non-teaching time for reference.
Source: https://www.oecd.org/content/dam/oecd/en/publications/reports/2015/02/how-much-time-do-teachers-spend-on-teaching-and-non-teaching-activities_g17a25d2/5js64kndz1f3-en.pdf
Across countries, teachers report spending an average of 38 hours per week on teaching and non-teaching activities, with about half of that time spent on teaching and the rest on planning, marking, collaboration, and admin.
When schools talk about “supporting teachers,” this is what they’re up against. AI agents can’t reduce teaching itself, nor should they. But they can reduce the drag of the non-teaching load.
What Even Is an AI Agent?
At this point, most of us have at least some experience with the basic chatbot. We get it: you type a question, it answers, you move on. An AI agent is different. It’s more like a digital assistant that can handle a chain of tasks within limits you set.
For a school, that might look like pulling information from approved systems, drafting a message, suggesting a plan, organising resources, or creating a summary, before handing it back to a human for a final check. The useful part isn’t that it knows things you don’t, but that it can help manage the fiddly, repetitive steps that swallow time.
If you’ve ever seen a teacher’s desktop (digital or physical), you’ll understand why this could be a game-changer.
An AI agent can’t calm a wobbly Year 3 who’s fallen out with their best friend. But it can reduce the admin that stops teachers from having the time to do exactly that.
For example, it can draft responses to common parent questions in the school’s tone, or turn a messy set of notes into a clean incident record for the staff to approve. It can summarise meeting notes into action points or help organise resources. The point is protecting energy.
Planning and Feedback
Lesson planning is one of those tasks that looks tidy from the outside and feels like juggling knives from the inside.
Used carefully, an agent can help teachers generate starting points with a rough lesson structure, a set of discussion prompts, or a differentiated worksheet draft that still needs a teacher’s judgement. The best versions of these tools would act like the assistant who lays everything out on the table so the teacher can make the real decisions.
Feedback is similar. Teachers know that specific, timely feedback helps learning, but writing it for thirty students, across multiple pieces of work, week after week, is intense. An agent can suggest comment phrasing aligned to a rubric, or create next-step prompts that a teacher then edits.
Supporting Students Means More Help and Fewer Gaps
This is the part that’s easy to get wrong in the conversation, because “AI in schools” instantly makes people picture a child alone with a screen.
But student support doesn’t have to mean replacing relationships. It can simply mean making the support teachers already provide more consistent and easier to deliver. An agent could help a teacher produce a quick recap for a student who was absent, generate practice questions tailored to what the class just covered, or translate classroom instructions for families who need language support, with proper review and policies in place, of course.
For older students, it might help with study planning by breaking a long assignment into steps or giving low-stakes practice questions that help them realise what they do and don’t understand. The teacher remains the master at work, while the agent just helps set up the “scaffolding” faster.
Early Years Deserve Extra Care
In early years settings, the best learning often looks like chaos to anyone who’s never tried it, with all the messy trays, imaginative play, and tiny arguments over who gets the blue shovel.
That’s why I’m cautious, yet hopeful, about how agents are used here. In my mind, the goal should be less screen time and less paperwork for educators, so they can spend more time guiding play-based learning in the way children actually need.
That might look like an agent helping with documentation, by turning brief observational notes into draft learning journal entries, or mapping observations to curriculum statements, so staff spend less time typing and more time interacting. The important thing is that the judgement about what matters, and what it means, has to stay human.
When School Isn’t in the Classroom
Parents know this one too well: illness, weather, transport issues, anxiety spikes, and family emergencies sometimes mean that just getting to school is hard enough.
In those moments, it’s all about keeping the thread of learning and belonging intact. Thoughtfully used, an agent could help teachers package the essentials (what we did, what to catch up on, what resources are needed) so students don’t fall behind during remote education periods, and so teachers aren’t rebuilding the same materials from scratch each time.
Closing Thoughts
If a school is considering AI agents, the best approach is to start small. Pick one workload pain point that staff agree drains time without improving learning, and keep the data minimal. Then, build clear approval steps and train staff on what the tool can and cannot do. Finally, measure whether it genuinely reduces workload and whether students experience any real benefit.
Most importantly, keep asking whether all this makes both students and teachers feel more supported.
Because if AI agents can help teachers spend less time buried in admin and more time guiding learning, noticing the quiet kid, celebrating progress, and building confidence (basically all the things they came into teaching for), then they’re a worthy inclusion.
Schools are short on time, and if we can give teachers even a little bit more of it, that feels like a future worth building.