Backwards Planning with Generative AI: Case Study Evidence from US K12 Teachers

Abstract

Problem definition: Generative AI technologies can help workers do tasks faster and better. However, many workers must plan which tasks to do, in addition to doing them. A prime example are K12 teachers, who plan their teaching tasks (i.e., activities, quizzes, projects) working backwards from end-of-year goals defined by state standards (a process known as backwards planning). In this paper, we ask: How are teachers, who must both plan and do tasks, beginning to use generative AI? Methodology/results: We conduct a longitudinal case study of 24 US public school teachers (all new to AI), sampled to vary by subject area and grade level. During the 2023–2024 school year, we gather from these teachers 360 minutes of recorded observation of generative AI use for their own work (not predefined by us) involving over 200 inputted prompts (and associated responses), together with 29 in-depth interviews and 34 generative AI use surveys. Analyzing this data corpus, we find that over the year teachers separate into three groups: (1) those who seek generative AI input (i.e., thoughts or ideas about learning plans) and output (i.e., quizzes, worksheets), (2) those who only seek generative AI outputs, and (3) those not using generative AI. The teachers in the first group—but not the second group—report productivity gains in terms of workload and work quality. Managerial implications: Workers can use generative AI for known tasks (task-level) or for deciding which tasks to do (workflow-level). Task-level use may save time completing tasks and increase the number of tasks done, but workflow-level use can save time deliberating about which tasks to do and nudge people toward the "right" tasks. AI developers, particularly in the education sector, ought to design tools that go beyond task-level use and nudge people toward more effective and efficient workflows. 

Keywords: generative AI, productivity, workflow, education operations, algorithm aversion