Conversational agents (CAs) are often unable to provide meaningful responses to user requests, thereby triggering user resistance and impairing the successful diffusion of CAs. Literature mostly focuses on improving CA responses but fails to address user resistance in the event of further response failures. Drawing on inoculation theory and the elaboration likelihood model, we examine how inoculation messages, as communication that seeks to prepare users for a possible response failure, can be used as an alleviation mechanism. We conducted a randomized experiment with 558 users, investigating how the performance level (high or low) and the linguistic form of the performance information (qualitative or quantitative) affected users’ decision to discontinue CA usage after a response failure. We found that inoculation messages indicating a low performance level alleviate the negative effects of CA response failures on discontinuance. However, quantitative performance level information exhibits this moderating effect on users’ central processing, while qualitative performance level information affected users’ peripheral processing. Extending studies that primarily discuss ex-post strategies, our results provide meaningful insights for practitioners.

Owing to Self-Service Business Intelligence (SSBI) systems' transformative power for organizations, substantial user uncertainties often blight their potential. Although these uncertainties pose a significant threat to effective SSBI implementation, their sources and determinants remain unclear. We conducted semi-structured interviews with 15 current users of a recently implemented SSBI system to empirically explore the relevant factors of user uncertainty. We undertook a rigorous thematic analysis of the collected data, thereafter developing a thematic map to visualize user uncertainties. This map uncovered three unexplored important factors (work routine change, social dynamics and fear of AI) for future research. Our findings show that users are not only perturbed by "hard" factors (e.g. a lack of technical understanding), but also by "soft" factors (social dynamics, fear of AI and nontransparency). Practitioners can use the thematic map to identify and observe potential uncertainties and to develop adequate procedures.