Companies automate the delivery of their online services by deploying artificial intelligence-based conversational agents (CAs). However, contemporary CAs still struggle to reliably answer the full range of requests from support seekers. To avoid service failure, service delivery activities of CAs and service employees should be interconnected by a handover of requests. This form of hybrid service delivery requires support seekers to disclose relevant information so that CAs can relay them to service employees prior to an imminent failure. By integrating and extending design knowledge from two DSR projects, we derive four design principles (DPs) to prepare handovers. These DPs guided the implementation of a service script in a CA prototype to facilitate the elicitation of information from support seekers. Based on two evaluation episodes, we show that support seekers feel supported by the CA in disclosing information which results in elaborate input for subsequent processing by service employees after handover.