Beyond AI Adoption in School Leadership: Task-Selective Integration Among Public Basic-School Headteachers in Ghana

Artificial intelligence (AI) is used in school leadership, but person-level adoption measures and task indicators conceal how realized use varies within leaders’ work. This study examines task-selective AI integration by separating task-use presence from conditional task-use frequency. Data were obtained from 187 public basic-school headteachers in three contexts in Ghana’s Ashanti Region who reported professional AI use during the previous six months. Responses to 11 leadership activities produced 2,057 headteacher-task observations. Mixed-effects logistic and cumulative-link ordinal models included task fixed effects and headteacher random intercepts. Task heterogeneity was evident for task-use presence (χ2(10) = 89.03, p < .001) and conditional frequency (χ2(10) = 151.21, p < .001). Reported information and communication technology (ICT)/digital training was positively associated with both margins. Reported AI-specific training was associated only with conditional frequency, while self-reported AI literacy did not independently differentiate either margin. Sensitivity analyses produced the same pattern. The findings show that post-adoption AI integration is not represented by an adopter label. Which leadership activities involve AI and how recurrently it is used once present constitute distinct dimensions of realized use. These findings support professional learning that combines digital-capability development with task-specific AI support, and implementation monitoring that looks beyond adopter status to where and how recurrently AI is used.