The preparation-time floor
Survey / self-reportMost of a teacher's working hours were not teaching. Planning, grading, parent communication and mandated paperwork consumed the majority of the week, largely unpaid and invisible, and were the primary driver of structurally predictable burnout.
LLMs absorbing the generation-and-drafting core of that work. Weekly AI users report saving 5.9 hours per week — six weeks per 37.4-week school year.
The Care work the Mode 1 narrative calls the substrate, resourced from time previously spent generating worksheets.
NurturingTeachingReinvesting reclaimed hours into the relational work that was always the real job.
ConnectingTeaching
- Gallup / Walton Family Foundation teacher survey population, 2025 [survey_self_report]
- Teachers report reinvesting saved time into nuanced feedback, individualised lessons, and getting home at a reasonable hour [survey_self_report]
Low risk on direction; moderate-to-high on magnitude; high on the distributional claim. That AI reduces preparation labour is solid. That it improves teachers' lives depends entirely on who captures the saving — a political question with no answer yet.
Time savings are self-reported by people motivated to believe the tool works; self-reported savings are systematically optimistic. More importantly, reclaimed time is not the teacher's property — if districts respond by increasing class sizes or adding duties, the six weeks evaporate and burnout returns with better worksheets. No mechanism guarantees the teacher keeps the gain.