Child helplines provide a platform for children facing serious challenges, allowing them to share their stories and receive emotional support and guidance through counselling sessions. To ensure that volunteers are well prepared for these interactions, effective training programmes are essential. Recently, a BDI-based conversational agent was developed to train counsellors through role-play simulations, with the agent taking on the role of the child. However, this simulation training does not provide guidance to the trainees that caused a decrease in the trainees’ self-efficacy. In this research, we aim to enhance the effectiveness of counselling training through role-play simulation by integrating a pedagogical agent. We integrated an adaptive pedagogical agent by applying the scaffolding technique, where we taught a set of skills divided into three modules. The pedagogical agent also provided feedback and hints as additional guidance methods. We evaluated this design through a mixed study involving 22 participants, comparing an intervention group trained with the pedagogical agent to a control group using a standard training approach. We measured the participants’ performance, self-efficacy, and perceived usefulness of the system. While the intervention group showed higher mean scores across all measures compared to the control group, the differences were not statistically significant, indicating a possible underpowered experiment. This study contributes to the integration of pedagogical agents in simulation training systems for child helplines by proposing a framework that combines scaffolding, adaptivity, and structured learning.