Designing Collaborative Intelligence Systems for Employee-AI Service Co-Production

Abstract

Employees increasingly co-produce services with artificial intelligence (AI). Focusing on system design, this research uncovers (1) which system features qualify an AI system as a so-called collaborative intelligence (CI) system, (2) to what extent CI systems influence work-related employee outcomes, and (3) which CI features relate to which outcomes. Based on an extensive literature review and a qualitative study, we demarcate CI from related concepts—such as hybrid intelligence, collective intelligence, and human-AI teaming—and identify five relevant CI system features: engagement, transparency, process control, outcome control, and reciprocal strength enhancement. Employing two scenario-based experiments with financial services employees (N = 309) and HR professionals (N = 345), we demonstrate that strong CI systems (i.e., characterized by the aforementioned five features) significantly relate to perceived service improvement, perceived outcome responsibility, (threat to) meaning of work, and adherence to the system. Particularly, transparency, process control, and outcome control are important design features, while, surprisingly, engagement seems less relevant. We also identify previous AI experience of employees as an important contingency factor: effects are much stronger for AI novices. Our research contributes to service literature by defining CI systems, while practitioners may benefit from our blueprint for CI system design.