Embedding AI in Lesson Planning

Evidence from a Multi-School Research Project

Abstract

This paper reports findings from a mixed-methods study examining the use of Artificial Intelligence (AI), including Generative AI (GenAI), to support lesson planning across five primary schools in England. Situated within contemporary Networked Learning scholarship, the study conceptualises AI-supported planning as a socio-technical practice shaped by professional relationships, institutional conditions, and collaborative learning cultures, rather than as a purely technical intervention. The research was conducted within a multi-academy trust engaged in digital transformation and committed to inclusive pedagogy through Universal Design for Learning (UDL).

Fifteen teachers across Early Years Foundation Stage, Key Stage 1, and Key Stage 2 participated over two school terms, piloting a range of AI tools including ChatGPT, TeachMate AI, Aila, Olex.AI, and Magma Maths. A convergent mixed-methods design was employed, combining baseline and follow-up teacher surveys, reflective journals, structured classroom observations, and pupil voice surveys and focus groups. Quantitative data captured changes in planning time, workload perceptions, and confidence, while qualitative data explored teacher sense-making, pedagogical decision-making, and learner experience.

The findings demonstrate a substantial reduction in average weekly planning time of 52.5%, decreasing from 10 hours to 4.75 hours, alongside an increase in teacher confidence in lesson planning from 50% to 100%. Qualitative analysis revealed that AI functioned most effectively as a “co-planner”, reducing the cognitive and emotional burden of planning while preserving professional judgement. Teachers reported improved curriculum alignment, greater pedagogical flexibility, and enhanced prompt literacy. AI-supported planning also enabled faster and more consistent production of differentiated materials, supporting inclusive practice for pupils with Special Educational Needs and Disabilities (SEND), English as an Additional Language (EAL), and those from disadvantaged backgrounds. Pupil voice data indicated increased engagement, independence, and enjoyment in AI-informed lessons.

The study also identified important tensions. While planning efficiency improved, time saved was often reallocated to other professional demands, highlighting a workload paradox that underscores the need for systemic policy responses. Analysis of the conditions underpinning successful adoption revealed the centrality of values-driven leadership, networked professional learning, psychological safety, reliable infrastructure, and ethical governance.

By combining quantitative workload data with rich qualitative insights, this study responds to calls within Networked Learning research for empirically grounded investigations of AI in educational practice. It demonstrates that AI’s educational value lies not in automation, but in its capacity to strengthen collaborative, reflective, and inclusive professional practice when embedded within trusted networks of learning.