Conversational Artificial Intelligence (CAI) has revolutionized human communication by enabling social and problem-solving conversations using natural language processing and generation technologies. However, while CAI-powered discussion platforms like D-Agree have been effective in promoting online discussions, general consensus facilitation policies have limitations in identifying the problems within proposed solutions. To address this issue, we propose a new facilitation policy in which the CAI interacts with the author of proposed solutions while raising the possibility of issues within the proposed solutions. We conduct experiments in a real-life setting in post-2021 Afghanistan, which demonstrate that the proposed approach is effective in collecting a higher number of objections raised for proposed solutions compared to the previous facilitation approach used in D-Agree.