Don’t Throw It Over the Fence! Toward Effective Handover from Conversational Agents to Service Employees

Abstract

Contemporary conversational agents (CAs) are capable of reliably answering repetitive low-complexity requests in online customer service, but regularly breakdown when dealing with high content or semantic complexity. The resulting service failures have a detrimental impact on customers’ satisfaction and their willingness to use CAs in the future. By aiming to avert CA breakdown in service encounters with a hybrid service recovery strategy via handover UI, we address a knowledge gap in service literature. As automated recovery strategies via conversation repair do not invariably prevent CA breakdown, real-time handover of customer interaction from CA to service employee (SE) is increasingly applied and investigated. This hybrid service recovery strategy places high demands on SEs, as they must keep waiting times short and avoid repetition of questions to customers after handover. Considering SEs limited cognitive capacities for information processing, we present a handover user interface (UI) with relevant information to support SEs after handover. Following a Design Science Research approach, we define design principles for the handover UI, based on meta-requirements derived from kernel theories and expert interviews. By evaluating the design principles via prototype instantiation, we show that the information types and their presentation in the handover UI keep cognitive efforts for SEs at a manageable level and help them initiate customer interaction quickly.