Where Artificial Intelligence Enters Teacher Work

Teacher use of artificial intelligence (AI) is increasingly common, but adoption does not show which parts of professional work become AI-supported. Using Teaching and Learning International Survey (TALIS) 2024 data, this study distinguishes AI adoption from task allocation and tests whether AI-using teachers are more likely to use AI for work they experience as relatively demanding. The adoption analysis included 56,669 teachers, and the primary allocation analysis included 24,058 AI users across 46 nonoverlapping education-system samples. Survey-weighted models matched AI use with demand in lesson planning, assessment/marking, and special-education support/adaptation. Mean focal-task demand was positively associated with AI adoption (OR = 1.08), and task demand relative to the same teacher’s focal-task mean was associated with task-specific AI use (OR = 1.16). The allocation relationship differed sharply by task (i.e., a two-point contrast in relative demand corresponded to a 3.2-percentage-point increase in predicted planning use, a 3.5-point decrease in assessment/marking use, and a 16.3-point increase in special-education support/adaptation). Cross-system heterogeneity was modest overall but more pronounced in task-specific profiles. These findings show that adoption alone provides an incomplete account of teacher AI integration. Where AI enters professional work depends on the task, and workload pressure does not produce a common pattern of use.