Acquiring a new skill, such as playing the piano, initially demands significant cognitive control. As one becomes proficient, reliance on cognitive control reduces, enabling skill automation. In adults, this transition is behaviourally evident through improved response times and accuracy, and neurally, by reduced frontal theta power across learning. However, our understanding of this transition throughout developmental stages is limited, especially concerning children and adolescents' ability to disengage cognitive control for skill automation. To address this gap, we measured neural activity with EEG while participants were performing a child-friendly task that required learning stimulus-response contingencies from response feedback across blocks of trials. Participants spanned three developmental stages: children (aged 6-7 years), adolescents (aged 12-14 years), and young adults (aged 18-24 years). A cluster-level permutation test was conducted for each group to identify specific time windows and channels of interest in theta power, which were subsequently analysed using linear mixed-effects models. Both age and the number of repetitions significantly impacted response time, with adolescents and adults exhibiting a steeper decrease across repetitions compared to children. Notably, despite successfully learning stimulus-response contingencies as evidenced by a significant decrease in error rate like other age groups, children's response times did not show a comparable decrease. In theta power, a similar age and practice effect was observed. The rate of decrease was steepest for adults, followed by adolescents, with children showing the least decrease. In sum, this research elucidates the differential reduction in cognitive control engagement during skill acquisition across different age groups.