When the prompting stops: exploring teachers' work around the educational frailties of generative AI tools

Teachers are now encouraged to use generative artificial intelligence (GenAI) tools to complete various school-related administrative tasks, with the promise of saving considerable amounts of time and effort. Drawing on interviews from 57 teachers across eight schools in Sweden and Australia, this paper explores teachers experiences when working with GenAI. In particular, it focuses on the large amounts of work that teachers put into reviewing, repairing and sometimes completely reworking AI-produced outputs that they perceive to be deficient. Rather than reflecting teachers lack of skill in prompting GenAI effectively, the paper shows how this work foregrounds the educational limitations and frailties of AI and other automated technologies with teachers having to act on a wide range of complex professional judgements around pedagogical appropriateness, social relations and overall educational value to bear on AI-generated content. The paper concludes by considering the need to challenge ongoing industry and policy claims around the labour-saving benefits of artificial intelligence in education, and instead focus on the ways in which these technologies are dependent on the hidden labour of humans to co-produce the illusion of automation.