Executive summary
England is losing tens of thousands of teachers a year. The Department for Education reports that 38,600 full-time equivalent teachers left the state-funded sector in 2024/25, equal to 8.5% of all qualified teachers. [1] DfE also reports that 57.7% of teachers who qualified ten years ago are still teaching in the state-funded sector, which means more than four in ten are not. [1] Education Support estimates that the current rate of teacher attrition costs the system over £1.5bn per year when training, recruitment, supply, productivity, workload and morale costs are taken into account. [2]
The causes are not mysterious. DfE evidence shows that among teachers and leaders considering leaving the English state school sector, high workload and stress and/or poor wellbeing are the two most commonly cited reasons, both reported by 89%. [3] Among those who actually left between 2024 and 2025, high workload remained the most common cited factor, reported by 74%. [3]
This paper argues that teacher retention should become one of the central measures by which AI in schools is judged. Not because AI can solve every reason teachers leave. It cannot. Pay, behaviour, SEND pressure, accountability, school funding and leadership culture all matter. But workload is one of the largest, most measurable and most addressable contributors to attrition.
This is not about asking teachers to produce more. It is about removing avoidable work so teachers have more time, headspace and professional energy for the parts of the job that only humans can do. By productivity, this paper means reducing duplicated preparation, unnecessary admin and workflow friction. It does not mean squeezing more output from already overloaded staff.
If technology can reduce the work that follows teachers home, protect their headspace, and give leaders evidence that workload is becoming more sustainable, it becomes more than a productivity tool. It becomes part of the retention infrastructure of the school system.
The risk is that AI arrives in the wrong form. A dozen disconnected AI tools may save minutes in one task while creating more logins, more checking, more policy burden, more data risk, and more work for teachers and leaders to stitch together. That is AI sprawl and it isn’t a strategy.
LESSO is being built around a different model. Teachers should review, adapt and approve work, not keep starting from a blank page. Leaders should be able to see adoption, workload impact, risk and policy compliance. AI should sit inside one governed school workflow, not across scattered tools where no one can see the whole. Sovereignty matters because the platform handling that workflow must be trustworthy, accountable and appropriate for sensitive school data.
The thesis is simple: if LESSO can prove sustained time saved, reduced take-home work, improved teacher headspace, repeat usage, SLT visibility and stronger intention to stay, then it is not merely an AI lesson-planning product. It is a retention and workload-sustainability platform for schools.
LESSO has not yet earned the right to claim long-term retention impact. It should be judged first on workload sustainability indicators: whether teachers save time, take less work home, feel more prepared, trust the outputs, and use the platform repeatedly inside real school workflows.
A note on perspective and evidence
I am not a teacher. I have never stood in front of a class for six hours, dealt with a safeguarding concern, marked books late at night, or tried to plan a week of lessons while also replying to parents and preparing for a staff meeting.
My view of teaching comes from proximity rather than direct experience. I am married to Amanda, a practising teacher, and I have watched the work follow her home for years. I have seen the difference between the part of teaching that gives energy and the part that drains it. The children were never the problem. The teaching was never the problem. The workflow around teaching is.
That outsider position matters. I have spent my career around systems, implementation and workflow. My instinct is to ask where the process is breaking, where work is duplicated, where context is lost, and why the person with the highest-value judgement is being asked to do the lowest-leverage work. In education, that person is often the teacher.
This whitepaper does not present original academic research. It assembles evidence from the Department for Education, Education Support, NFER, the House of Commons Library, Gallup and Walton Family Foundation, and published research on teacher-student relationships. The argument is a synthesis. It is intended to be tested, challenged and improved by people who know the system better than I do.
Where the evidence is strong, the paper says so. Where the claim is a thesis, it is framed as a thesis. Where LESSO still has to prove its impact, the paper says exactly what should be measured.
1. The retention crisis is now a productivity crisis
Teacher retention is usually discussed as a workforce problem. That is correct, but incomplete. It is also a productivity problem, a quality problem, a public value problem and a human problem.
The latest DfE School Workforce data shows that 38,600 full-time equivalent teachers left the state-funded sector in 2024/25. [1] That was an improvement on the previous year, but it still represents one in twelve qualified teachers. The same release reports that 91% of leavers left the state-funded sector rather than retiring. [1] In other words, this is not only a normal retirement cycle. It is experienced teachers leaving the system, changing career, moving into other education sectors, reducing hours, or otherwise stepping away from the roles schools need them to hold.
The long-term retention picture makes the problem clearer. DfE reports that 89.7% of teachers who qualified in 2024 were still teaching one year later, 75.8% of those who qualified three years ago were still teaching, 67.4% of those who qualified five years ago were still teaching, and 57.7% of those who qualified ten years ago were still teaching in the state-funded sector. [1] Every lost teacher represents public investment, school knowledge, curriculum expertise, relationships with pupils, and operational capacity that has to be rebuilt.
Education Support puts an economic frame around that loss. In 2023/24, it says 37,021 teachers left the profession for reasons other than retirement. [2] It estimates that the cost of training those teachers was over £1bn in real terms. [2] When additional turnover costs are included, including lost productivity, recruitment, supply, workload and morale impact on the wider team, Education Support estimates that the cost of the current rate of attrition rises to over £1.5bn per annum. [2]
That is the economic number this paper is built around, but it is not the whole story.
Not because it captures every cost. It does not. It does not fully price the effect of staff churn on pupil confidence, curriculum continuity, parental trust, behaviour culture, mentoring, middle leadership capacity or the remaining staff who absorb the pressure. It is useful precisely because it is conservative enough to be credible and large enough to be impossible to ignore.
If Education Support estimates that attrition is costing the system over £1.5bn per year, then the question for any serious education technology company is not only whether its product looks impressive. The question is whether it attacks any part of the avoidable workload that contributes to that leakage.
2. Workload is one of the largest holes in the bucket
The reasons teachers leave are complex. No serious person should pretend that one product, policy or intervention can solve teacher retention on its own. Pay matters. Behaviour matters. SEND pressure matters. Leadership culture matters. Inspection pressure matters. Funding matters. Flexible working matters. The public perception of teaching matters.
But workload is consistently one of the clearest and most addressable drivers.
DfE’s Working Lives of Teachers and Leaders Wave 4 report found that, among teachers and leaders considering leaving the English state school sector, high workload and stress and/or poor wellbeing were the two most commonly cited reasons, both reported by 89%. [3] The same DfE report found that among actual leavers between 2024 and 2025, high workload, stress and/or poor wellbeing, and teachers’ views not being valued by policymakers were the top three reasons cited as important in contributing to their intention to leave. High workload remained the most common factor, cited by 74% of leavers. [3]
NFER’s 2025 Teacher Labour Market report reaches the same direction from a different evidence base. [4] It states that 90% of teachers considering leaving teaching in 2023/24 cited high workload as a factor, and that pupil behaviour has become one of the fastest-growing contributors to workload since the pandemic. [4]
This matters for the AI conversation because workload is not an abstract complaint. It is a set of repeated, observable tasks that fill the gaps around teaching: planning, adaptation, marking, feedback, parent communication, resource creation, reporting, assessment, admin, meetings, behaviour notes, safeguarding records and the mental load of knowing tomorrow is not ready yet.
The crucial point is that the work does not arrive in one neat place. It leaks across the week. A teacher teaches, then plans, then checks email, then adapts work for pupils with different needs, then writes a parent message, then marks, then thinks about Friday’s behaviour note, then remembers that next week’s resources still need to be made. That is why workload is not only about hours. It is about headspace.
A teacher can survive a long week. What erodes people is a long week that never properly ends.
That is the part AI should be judged against. Does it reduce the work that follows teachers home? Does it change the first action from creation to review? Does it make the week feel more prepared before it begins? Does it protect the teacher’s professional judgement while removing the blank page? Does it give leaders evidence that workload is moving in the right direction?
If it does not, it may still be a useful tool. But it is not a retention intervention.
3. Retention is not separate from pupil outcomes
It would be easy to frame this as a teacher wellbeing paper only. That would be too narrow.
Education Support is clear that teacher retention is a cornerstone of educational quality and equity. [2] It argues that retaining experienced, skilled teachers gives children and young people greater continuity, stronger relationships and more effective teaching, while high turnover disrupts learning environments, undermines progress and disproportionately affects pupils in disadvantaged communities.
That is not just common sense. It is supported by wider research on the importance of teacher-student relationships. A systematic second-order meta-analytic review synthesised more than 70 years of educational research across 26 meta-analyses, covering approximately 2.64 million pre-kindergarten and K-12 students. [9] It found significant relationships between teacher-student relationships and a range of outcomes, including academic achievement, motivation, executive function, behaviour, school belonging and wellbeing.
The implication is not that AI improves pupil outcomes by itself. That claim would be premature and weak. The stronger claim is that teachers are part of the learning infrastructure. Stable, skilled, less exhausted teachers are better positioned to notice pupils, build trust, adapt teaching, respond with judgement, and create the human connection that children need.
This is why the workload conversation matters. A teacher who spends Sunday evening creating slides from scratch is not only losing time. They are losing recovery, focus and emotional capacity. A teacher who spends Monday morning stitching together five systems before the first pupil walks in is starting the day with friction rather than attention. A teacher who is constantly behind is less able to be fully present.
In that sense, teacher workload is not a back-office issue. It is upstream of the classroom.
The point of AI in schools should not be to replace the relationship between teacher and pupil. It should be to protect it.
That gives us a different test for education AI:
Does it give teachers more capacity for the human part of the job?
If the answer is yes, it deserves attention. If the answer is no, it may simply be another layer of digital work.
4. AI can help, but only if it changes the workflow
There is now enough evidence to say that AI can save teachers time. The question is not whether AI has any utility. It does. The question is whether schools can turn isolated utility into sustained, governed, measurable workload reduction.
Gallup and the Walton Family Foundation found that teachers who use AI tools at least weekly estimate saving an average of 5.9 hours per week, equivalent to roughly six weeks across a school year. [6] The same research found that teachers report time savings across tasks such as worksheets and assessments, administrative work and preparation to teach. It is US-based research, so it should not be transplanted into England without caution, but it gives a useful signal: when teachers use AI regularly and well, the time dividend can be material.
The Department for Education’s own framing is also moving in this direction. Its generative AI policy paper states that generative AI has demonstrated that it can help the education workforce by reducing some of the administrative burdens teachers, staff and school leaders face. [7] DfE guidance for schools also says teachers can use AI to help with planning lessons, creating resources, marking work, giving feedback and handling administrative tasks, provided they use professional judgement and check that outputs are accurate and appropriate. [8]
The opportunity is real.
The danger is that schools mistake access to AI for an AI operating model.
A teacher using ChatGPT to draft a worksheet may save time. Another using Copilot to draft an email may save time. Another using a marking assistant may save time. Another using a slide generator may save time. Each decision is understandable. Teachers are overloaded, and relief is valuable wherever it can be found.
But across a school or trust, the picture changes. One teacher uses one tool. Another uses another. A third copies pupil context into a platform the school has not approved. A fourth gets a good output but has to reformat it, check it, move it, explain it and recreate it somewhere else. SLT does not know who is using what, what data is being entered, whether policy is being followed, whether workload is genuinely falling, or whether the school has simply moved work into a dozen invisible places.
That is why I’ve called this AI sprawl.
AI sprawl is what happens when useful tools accumulate faster than governance, workflow design and measurement. It may feel like innovation at classroom level while becoming risk at school level.
The lesson from the last decade of EdTech should be clear. More tools do not automatically create better systems. In many cases, they create more integration work for teachers and more oversight burden for leaders.
AI will only become a retention lever if it does more than generate content. It has to change the shape of the work.
5. The missing layer is governed workflow
The current debate often treats teacher workload, AI governance and data sovereignty as separate issues. In practice, they are becoming the same issue.
Workload is the pressure that makes teachers reach for AI.
Governance is the thing school leaders need once teachers begin using AI.
Sovereignty is the trust condition that matters when sensitive school data starts moving through AI systems.
A school cannot solve those problems with a loose collection of disconnected tools. It needs a governed workflow layer.
That layer should do six things.
First, it should reduce workload at source. The teacher should not have to start from a blank page every time. Planning, resources, communications and assessment support should be prepared from the school’s context, ready for teacher review.
Second, it should keep the teacher in control. AI should draft, organise, suggest and prepare. The teacher should review, adapt and approve. Wherever professional judgement about a child is involved, the human decision point should be explicit.
Third, it should carry context forward. A lesson plan should inform the slides. The slides should inform the task. The task should inform assessment. The assessment should inform next week’s teaching. If the teacher has to keep carrying context manually between tools, the platform has not solved the workflow.
Fourth, it should give leaders visibility. SLT should be able to see adoption, repeat usage, hours returned, outputs generated, policy compliance, audit trails and risk signals. Without visibility, leaders cannot govern AI adoption or prove impact.
Fifth, it should reduce the number of places teachers need to go. A platform should not be another tab on top of everything else. It should become the place where routine work is prepared, reviewed and approved.
Sixth, it should treat data governance as structural rather than decorative. If the platform is handling lesson context, pupil work, parent communication, assessment or school policy, then data location, processing, supplier control and legal jurisdiction matter.
This is the distinction between a tool and a platform.
A tool helps with a task.
A platform changes how the work connects.
That distinction is central to LESSO.
6. LESSO’s thesis: retention and productivity, powered by governed AI
LESSO is being built as a workload-sustainability and retention platform for schools, powered by governed AI.
That is a deliberate repositioning.
LESSO is not just an AI lesson planner. Lesson planning is one workload surface. It is not the whole problem.
LESSO is not just a chatbot. A chatbot can be useful, but the school day does not live inside a blank chat box.
LESSO is not just a content generator. Generating slides, worksheets or plans may be valuable, but if those outputs sit outside the school’s workflow, the teacher still becomes the integration layer.
The product direction is different: one trusted platform where teachers prepare, review and approve the work that otherwise follows them home, and where leaders can see whether AI adoption is reducing workload safely and consistently. This is productivity in the educational sense: less duplicated effort, less reformatting, less copying between systems and less avoidable administration, not an expectation that teachers simply produce more.
At the teacher level, this means moving from creation to review. The teacher provides context: year group, topic, curriculum, class needs, school expectations and professional judgement. Mrs J, the LESSO co-teacher, prepares the first version of the work: lesson materials, resources, communication, differentiation support, assessment support and next steps. The teacher reviews, edits and approves. The judgement stays with the professional. The blank page is removed.
In practice, this means a teacher can ask what needs preparing this week, generate lesson materials from school context, create supporting resources, draft parent communication, review assessment support and keep the work connected rather than moving between disconnected systems. The platform should not replace the teacher. It should reduce the avoidable work around the teacher.
At the leadership level, this means moving from invisible AI use to governed adoption. Leaders should not have to guess which tools staff are using, where data is going, or whether workload is improving. They should be able to see usage, adoption, workflow impact, audit trails and policy alignment in one place.
At the school system level, this means moving from AI sprawl to one governed operating model. AI should not become another wave of fragmented EdTech that creates procurement risk, training burden and inconsistent practice. It should become a controlled layer that supports teacher workload, leadership visibility and school trust.
Sovereignty sits inside this thesis, but it is not the first sentence. The first sentence is workload. The reason sovereignty matters is that any platform serious enough to reduce workload across real school workflows will inevitably handle sensitive context. If the platform is helping with assessment, parent communication, pupil work, policy compliance or school-level reporting, then the school needs to know who controls the data, where it is processed, which law applies, who could compel access, and what happens if a supplier changes, withdraws or restricts the service. In school language, sovereignty means practical control: approved use cases, clear data processing, role-based access, auditability, human review points and accountability.
In other words, sovereignty is not a separate marketing claim. It is part of the trust architecture of a retention platform.
7. What LESSO must prove
The argument in this paper should not be accepted because it sounds plausible. It should be tested.
LESSO should be judged by whether it can move measurable upstream indicators of retention.
It is unrealistic to expect a six-week pilot to prove that a teacher stayed in the profession three years later because of one platform. That would be a weak claim and an unfair standard. But a pilot can test whether LESSO changes the conditions that evidence links to retention: workload, stress, headspace, take-home work and the sustainability of the job.
The pilot model should therefore measure the following.
Hours returned per teacher per week.
How much time does LESSO save across planning, preparation, communication, assessment and admin? This should be measured through weekly teacher self-reporting, usage data and task-level comparison where possible.
Reduction in take-home work.
Does LESSO reduce the number of evenings or weekend blocks spent preparing school work? The most important question is not only “how many minutes were saved?” but “did this change whether the work followed you home?”
Teacher headspace.
Teachers should be asked whether their workload feels more manageable, whether they feel more prepared for the next day, whether they can switch off more easily, and whether the platform reduces the mental load of preparation.
Repeat usage.
A demo proves curiosity. Repeat usage proves value. Weekly active usage, repeated workflow completion and return behaviour matter more than one-off output generation.
Quality and trust.
Teachers should rate whether outputs are accurate, useful, editable and aligned with their class needs. The goal is not fully autonomous teaching. The goal is a trustworthy first version that a teacher can adapt faster than they could create from scratch.
SLT visibility.
Leaders should be able to see adoption rates, active users, workflows completed, hours returned, policy compliance and risk signals. Without leadership visibility, the product may help individual teachers but fail as a school platform.
Retention intent and sustainability.
Teachers should be asked whether LESSO makes their workload feel more sustainable, whether it reduces stress linked to preparation and admin, and whether a platform like this would make them more likely to stay in their role or school.
The key is not to claim too much too early.
A credible pilot should not say, “LESSO improves pupil outcomes.” It should say, “LESSO reduced workload by X, reduced take-home work by Y, improved teacher-reported headspace by Z, achieved repeat usage across A% of teachers, and gave leaders visibility of adoption and impact.”
That is the bridge from product to retention thesis.
If LESSO can prove those measures repeatedly, across schools and over time, then the investor story changes. This is not another lesson-planning app fighting for attention in a crowded market. It is a platform testing whether governed AI can reduce the avoidable workload that contributes to England’s teacher attrition problem.
8. How school and trust leaders should evaluate AI platforms
Every AI vendor in education will claim to reduce workload. Many will also claim to be safe, governed, compliant or school-ready. Leaders need questions that cut through those claims.
Here are the questions I would ask any vendor, including LESSO.
What workload does the product actually reduce?
Ask the vendor to name the specific tasks. Planning is not enough. Does it reduce preparation, resources, assessment, communication, differentiation, admin, reporting or leadership oversight? Where exactly is the time saved?
How is the time saving measured?
A vendor should not rely only on anecdotes. Ask what will be measured before, during and after a pilot. Ask whether they report hours returned, take-home work reduction, repeat usage and teacher confidence.
Does the product change the workflow, or just generate outputs?
A generator gives the teacher something to copy, check and move elsewhere. A platform should carry context across the workflow and reduce the need to stitch tools together manually.
Where does the teacher approve?
Professional judgement should remain with the teacher. Ask where human review sits before any output is used with pupils, parents or records.
What can SLT see?
If leaders cannot see adoption, impact, risk and policy compliance, they cannot govern the platform. Visibility is not a nice-to-have. It is the difference between individual experimentation and school strategy.
What data is processed, where and under whose control?
UK data residency and true sovereignty are not always the same thing. Ask where data is stored, where AI processing happens, which providers are involved, which legal jurisdictions apply, and who could compel access or service changes.
What is live today and what is roadmap?
This is one of the most important questions in AI procurement. Vendors are often selling the future. A roadmap is not a capability. Schools should ask for a written distinction between what works now, what is in pilot, what is planned and what depends on third-party integration.
What happens if the vendor disappears, changes provider or loses access to a model?
AI capability is increasingly dependent on infrastructure, model access and geopolitical decisions. Schools should know how resilient the service is and what dependencies sit underneath it.
These questions are not meant to stifle innovation in the sector but they are the minimum standard for serious adoption.
If AI is going to touch teacher workload, pupil context and school decision-making, it should be held to the same seriousness as any other part of school infrastructure.
9. Recommendations
A whitepaper is only useful if it changes what someone does next. The recommendations below are deliberately practical.
For school leaders
Start by treating AI adoption as a workload and retention question, not only a technology question. Ask where staff are already using AI, which tasks they are using it for, what data they are entering, and whether it is actually reducing work or just shifting work into new places.
Map the five highest-friction teacher workflows in your school: planning, resource creation, assessment, parent communication, and reporting or admin. Then ask which of those workflows could move from creation to review.
Do not buy AI on demo quality alone. Ask vendors to define workload metrics before the pilot begins. Require reporting on hours returned, repeat usage, teacher confidence, policy compliance and take-home work.
For multi-academy trusts
Treat AI sprawl as a trust-level governance risk. The problem is not that teachers are experimenting. The problem is that experimentation becomes ungovernable if every school, department and teacher is using a different tool with different data flows and different controls.
Set a trust-wide AI operating model. Define approved use cases, approved platforms, data rules, review points, audit requirements and impact measures. If workload reduction is the aim, require vendors to prove it in a way that can be compared across schools.
Look for platforms that give central visibility without removing professional judgement from teachers. The best architecture is not command and control. It is governed autonomy.
For local authorities and education system leaders
Make teacher retention and workload impact explicit in AI evaluation. The question should not simply be whether AI tools are safe or innovative. It should be whether they reduce the workload drivers that push teachers out of the profession.
Encourage schools to measure upstream retention indicators. Hours returned, reduced take-home work, improved headspace, repeat usage and intention to stay are all practical measures that can be captured before long-term retention outcomes are visible.
Support a common language around AI procurement. Schools need to distinguish between UK-hosted, UK data residency, and true operational sovereignty. They also need to distinguish between a task tool and a governed workflow platform.
For policymakers
Connect the teacher retention strategy to the AI strategy. The education system should not evaluate AI mainly by novelty, adoption or output volume. It should evaluate AI by whether it improves staff productivity, wellbeing, retention and the quality of teacher-pupil interaction.
Use testbeds and pilots to measure the right things. The most important question is not “can this AI generate a lesson?” The better question is “does this AI reduce the work that drives teachers out of the profession, while keeping professional judgement and school governance intact?”
10. Conclusion: judge AI by whether it helps teachers stay
The future of AI in schools should not be measured only by what it generates. It should also be measured by whether it helps teachers stay.
That is the core argument of this paper.
England is losing tens of thousands of teachers a year. Education Support estimates that the annual cost of teacher attrition is over £1.5bn. Workload and stress are among the clearest drivers. Teacher retention matters not only because recruitment is expensive, but because pupils need stable, skilled, present adults who know them.
AI can help, but only if it is built and adopted in the right form.
A scattered collection of tools will not solve workload if teachers still have to stitch everything together. A chatbot will not solve retention if it creates another place to start work. A content generator will not solve governance if leaders cannot see what is happening. A platform will not earn trust if schools cannot understand where the data goes or who controls it.
The opportunity is larger than lesson planning.
The opportunity is to build a governed workflow layer for schools that gives teachers time back, gives leaders visibility, reduces AI sprawl, and makes workload reduction measurable.
That is the direction LESSO is being built towards.
Not more tools for teachers to manage.
One trusted platform that helps schools close the workload leak.
If you lead a school, trust, local authority, education body or policy organisation, I would welcome challenge on the thesis. The argument should be tested before it is accepted. That is why I am putting it in writing.
About LESSO
LESSO is a UK company building a governed AI platform for schools. Its purpose is to reduce teacher workload, improve teacher headspace and give leaders the visibility they need to adopt AI safely, consistently and measurably.
LESSO is built around Mrs J, an AI co-teacher and agentic assistant designed to support the work that surrounds teaching. Mrs J is not intended to replace professional judgement. She prepares work for review, refinement and approval by the teacher.
The platform has launched with an AI agent chat experience, a multi-lesson wizard, document-aware planning workflows, assessment support, Microsoft 365 integration, a school leadership suite and LESSO Academy.
The multi-lesson wizard allows teachers to create full lesson bundles from school documents such as knowledge organisers, curriculum plans and schemes of work. A teacher can generate a sequence of lessons, including slides, lesson plans, activities and notes, in under ten minutes. The workflow is designed to support full units of work, not isolated one-off outputs, and can be scheduled so preparation can take place autonomously ahead of the school day.
LESSO also includes assessment workflows backed by mark schemes and exam-board-aligned criteria, including GCSE awarding body contexts. The aim is not simply to mark work faster, but to help teachers identify patterns, misconceptions and next steps with less manual effort.
Through Microsoft 365 integration, Mrs J can work with the documents and communication channels schools already use. This includes sourcing, reading, embedding and recalling relevant OneDrive documents, supporting email drafting, helping prioritise inboxes and connecting school context to the work being prepared. LESSO’s approach is pragmatic: schools already rely heavily on Microsoft, so the platform must work with the current operating reality while building towards a more sovereign future.
The SLT Suite is designed to give leaders the governance layer that individual AI tools cannot provide. This includes visibility over adoption, staff use, school setup, policy alignment, communication workflows and the indicators needed to understand whether AI is reducing workload or simply adding another layer of technology.
LESSO Academy will support the wider adoption challenge by providing training, CPD, AI guidance, product education, sector updates and future thought leadership content. The platform is not only a tool for producing content. It is intended to help schools build confidence, capability and governance around AI.
The longer-term roadmap is more ambitious. LESSO is being built towards a sovereign education stack: a school-controlled platform that can eventually support core productivity workflows, pupil-facing AI under strict safeguards, and a UK sovereign education AI model trained and hosted for education-specific use cases.
This roadmap matters because the end state is not simply better lesson generation. The end state is school-controlled AI infrastructure: one trusted platform where teacher workload, leadership visibility, governance, assessment, communication and future pupil support can connect safely.
LESSO’s position is that schools should not have to choose between innovation and control. They need AI that gives teachers time back, helps leaders govern adoption and keeps the educational workflow under school and UK jurisdiction.
That is the journey LESSO is building towards.
Contact: luke@lesso.co.uk
References
- [1] Department for Education. School workforce in England: November 2025. Published 4 June 2026. https://explore-education-statistics.service.gov.uk/find-statistics/school-workforce-in-england/2025
- [2] Education Support. Revisiting the teacher retention crisis: recommendations for change. Autumn 2025. https://www.educationsupport.org.uk/media/qawn3vdn/revisiting-the-teacher-retention-crisis-recommendations-for-change.pdf
- [3] Department for Education. Working lives of teachers and leaders: wave 4 summary report. Published 27 November 2025. https://assets.publishing.service.gov.uk/media/693bdc426a12691d48491e9b/Working_lives_of_teachers_and_leaders_wave_4_summary_report.pdf
- [4] National Foundation for Educational Research. Teacher Labour Market in England Annual Report 2025. Published 13 March 2025. https://www.nfer.ac.uk/publications/teacher-labour-market-in-england-annual-report-2025/
- [5] House of Commons Library. Teacher recruitment and retention in England. Updated 2 February 2026. https://commonslibrary.parliament.uk/research-briefings/cbp-7222/
- [6] Gallup and Walton Family Foundation. Teaching for Tomorrow: Unlocking Six Weeks a Year With AI. Published June 2025. https://static.waltonfamilyfoundation.org/df/fb/eba12807470a9402d7433cc47dba/teaching-for-tomorrow-unlocking-six-weeks-a-year-with-ai-report.pdf
- [7] Department for Education. Generative artificial intelligence (AI) in education. Updated 12 August 2025. https://www.gov.uk/government/publications/generative-artificial-intelligence-in-education/generative-artificial-intelligence-ai-in-education
- [8] Department for Education Education Hub. AI in schools and colleges: what you need to know. Published 10 June 2025. https://educationhub.blog.gov.uk/2025/06/artificial-intelligence-in-schools-everything-you-need-to-know/
- [9] Emslander, V., Holzberger, D., Ofstad, S. B., Fischbach, A., and Scherer, R. Teacher-student relationships and student outcomes: A systematic second-order meta-analytic review. Psychological Bulletin, 2025. Summary indexed at PubMed: https://pubmed.ncbi.nlm.nih.gov/39928458/ and OSF overview: https://osf.io/5de6p/overview