Research reveals that in highly structured learning environments pedagogical agents can act as tools to direct students' learning processes by providing content or problem solving guidance. It has not yet been addressed whether pedagogical agents have a similar impact in more open learning environments that aim at fostering students' acquisition of complex problem solving skills. To fill this gap, this study compares two experimental conditions to a control condition. In the first experimental condition, the pedagogical agent's interventions are adapted to students' activities (APA-group). In the second experimental group, the agent interferes on fixed time intervals (NPA-group). Students work on an environmental problem in an open learning environment. It was hypothesized that the two experimental groups would outperform the control group and, additionally, that the APA-group would outperform the NPA-group. Results show that the NPA-group outperforms the control group, and that the APA-group takes a middle position, not differing from either the NPA-group or the control-group. Consequently, the second hypothesis was not confirmed. An agent providing adapted advice did not result in better performance. The discussion provides some possible explanations for these results. (Contains 1 table and 2 figures.)