Appearance, voice features, and communication style affect users trust in conversational agents (chatbots), but few studies have assessed what features users like and dislike. Using design thinking, we developed Susa, a conversational agent, to help workplaces promote teamwork and collaborative practices. Design thinking prioritizes co-creation and multidisciplinary teamwork to develop innovative solutions to complex problems. The aim of this qualitative study was to explore users' interactions

more »

... with and reactions toward Susa and explain how we used user inputs to adapt and refine the first prototype. The employees and managers from four workplaces participated in three workshops to test and refine the agent. We applied an explorative thematic analysis of data collected via video recordings of the workshops. The results of the analyses revealed that visual identity, communication style and personality was important for acceptability. Users favored a more human like agent that primarily communicated with the team via text messages. Users disliked emoticons and humor because these features clashed with the seriousness of the topic. Finally, users highlighted that Susa helped structure organizational change processes, develop concrete action plans, and stay on track. It is a weakness that Susa is a simple robot based on a preprogrammed script that does not allow users to adapt the process.