By Dr Haris Mughal, Running Paper. Drawing on an independent literature evaluation by Dr Fiona Aubrey-Smith.
Ask any teacher what eats into their evenings and marking will be near the top of the list. It matters. It is also one of the clearest examples of how workload can contribute to exhaustion and burnout.
This is the first in a short series drawing on an independent literature evaluation of Running Paper by Dr Fiona Aubrey-Smith. Each post looks at what published research says about a challenge schools face. We start with one of the biggest: teacher workload.
Key takeaways:
- Repetitive marking and data entry are what researchers call hindrance demands, the kind of workload consistently linked to exhaustion.
- In theory, AI can act as a resource that takes on this work and frees teachers for the parts of the job that motivate them.
- None of it works without trust. If teachers feel they must check every mark, the time saved disappears.
Burnout comes from an imbalance
One of the most widely used frameworks in occupational psychology is the Job Demands Resources model (Demerouti et al., 2001). Every job has demands: the parts that take sustained effort and carry a physical or mental cost. Every job also has resources: the things that help you get the work done, ease the demands or help you grow.
When demands outweigh resources over time, the risk of exhaustion and burnout rises. When resources keep pace, people are far better placed to cope.
Some workload costs more than it gives
Researchers draw a useful line between two kinds of demand.
Challenge demands stretch you. They are hard but they can build mastery, like planning a new unit or taking on a tricky class.
Hindrance demands take effort without giving much back. Repetitive admin, data entry and high volume grading are typical examples. These are the demands consistently linked to exhaustion, disengagement and teachers leaving the profession (Skaalvik and Skaalvik, 2020).
Getting feedback to students is valuable. The hours spent getting there often are not. Dr Aubrey-Smith’s evaluation notes that deep marking policies, long common in UK schools, have added to this and can push workload beyond what many teachers can sustain (NEU, 2023).
Marking quietly drains mental energy
Cognitive Load Theory (Sweller, 1988) is usually applied to how students learn. Recent work has extended it to how teachers work (Rind, 2026).
Teaching already carries a heavy load. After school marking adds another layer on top:
- deciphering handwriting
- working out what a student meant
- copying scores into spreadsheets, markbooks and the MIS
The argument is that all of this competes for the mental energy teachers need for the work that moves learning forward: reflecting on what went wrong and planning what to do next. A teacher also has to gather the data before they can analyse it. By the time the spreadsheet is finished, much of the energy for thinking about what it means has gone.
AI could give that energy back
Some recent research has started to treat AI in schools as more than a productivity tool. It is beginning to be framed as part of protecting teachers’ wellbeing (Yadav and Hamid, 2025).
The theory is simple. If AI takes on some of the hindrance demands, teachers can put their energy back into the relational and creative parts of the job. Mentoring. Curriculum design. Getting to know their students. These are intrinsic sources of motivation and professional satisfaction for teachers. (Zhang and Cao, 2025).
Trust decides whether the time saving is real
Unreliable technology does not remove work. It creates new work. As Dr Aubrey-Smith puts it, “trust is therefore the operational currency of such systems”.
Research on technostress warns that if people feel they have to double check everything a tool produces, the time saved disappears (Högemann et al., 2025). A tool that saves an hour but leaves a teacher anxious about every mark has not reduced the demand. It has swapped one kind for another.
Two other points are worth being upfront about. Freeing up time only helps if that time goes somewhere useful, which depends on the school as much as on any tool. This kind of support also fits best where schools already use tests and practice papers regularly. That is one approach to assessment. It is not the only good one.
Where Running Paper fits
Running Paper was built around this problem. Students still work on paper, which remains central to assessment in UK schools and exams. Running Paper reads and marks the handwritten work, flags anything it is unsure about for the teacher to check and builds the question level analysis automatically. In the language of the research, the aim is to act as a job resource.
Dr Aubrey-Smith’s evaluation concluded that, in theory and if implemented effectively, a tool designed this way could act as a genuine protective factor against teacher burnout in schools that already use tests and practice papers regularly. She also identified trust as a critical hurdle.
That is why every answer is marked by two independent AI models, with any answer where they disagree beyond a set tolerance flagged for the teacher to review. Teachers always have the final say. The platform’s job is to take on the repetitive work, so teachers can put their judgement where it matters most.
Next in the series
Can AI really read a student’s handwriting? We look at how accurate it has become, where it falls short and why that matters.
About this series: these posts draw on an independent literature evaluation by Dr Fiona Aubrey-Smith, produced for the EdTech Evidence Board pilot. The pilot was delivered by the Chartered College of Teaching and funded by the Department for Education to evaluate the evidence behind EdTech products. You can read the full report here.
References
Demerouti, E., Bakker, A.B., Nachreiner, F. and Schaufeli, W.B. (2001) ‘The job demands-resources model of burnout’, Journal of Applied Psychology, 86(3), pp. 499–512.
Högemann, M., Hein, L., Britsche, J.-O. and Thomas, O. (2025) ‘Technostress and generative AI in the workplace: a qualitative analysis of young professionals’, Frontiers in Artificial Intelligence, 8, article 1728881.
NEU (2023) The Impacts of Ofsted Inspection on Schools and Teachers. London: National Education Union.
Rind, I.A. (2026) ‘Conceptualizing the impact of AI on teacher knowledge and expertise: a cognitive load perspective’, Education Sciences, 16(1), article 57.
Skaalvik, E.M. and Skaalvik, S. (2020) ‘Teacher burnout: relations between dimensions of burnout, perceived school context, job satisfaction and motivation for teaching. A longitudinal study’, Teachers and Teaching, 26(7–8), pp. 602–616.
Sweller, J. (1988) ‘Cognitive load during problem solving: effects on learning’, Cognitive Science, 12(2), pp. 257–285.
Yadav, V. and Hamid, M. (2025) ‘The impact of artificial intelligence on teacher burnout and professional well-being’, International Journal of Research & Technology, 13(4), pp. 638–649.
Zhang, H. and Cao, J. (2025) ‘From digital disruption to mental health: the impact of AI-induced educational anxiety on teacher well-being in the era of smart education’, BMC Public Health, 25(1), 4010.
