Drivers and Barriers of Generative and Conversational Artificial Intelligence for the Workplace: Organizational and Design Considerations

Artificial Intelligence (AI) as a technology is foundational to the majority of software applications today. In particular, with the emergence of ChatGPT in late 2022—a conversational agent based on a large language model (LLM)—generative AI (GenAI) became the new standard for AI. With its power to consume, analyze, process, and finally generate data, potential use cases are numerous. For many organizations, GenAI is the future technology due to its potential to increase efficiency and productivity through automated analyses and predictions, while decreasing costs and human labor for better scalability. As individuals, GenAI tools have also become the norm in our private lives, whether it be talking to powerful and seemingly all-knowing assistants, retrieving information in a digestible way, or generating texts for different purposes. However, using GenAI tools is also perceived critically by many. Disadvantages include hallucinations, monetary costs, ethical concerns, and environmental implications associated with GenAI usage, and the fact that data privacy may not always be provided. Consequently, for a sensitive environment like the workplace, it is a fine balance between benefits and risks. The aim of this dissertation is to shed light on drivers and barriers of GenAI usage at the workplace with a focus on knowledge workers. Knowledge workers generate value in their jobs by understanding, transforming, and transferring knowledge and information. As the core capabilities of GenAI exactly align with these types of tasks, they represent the group of workers whose work provides the largest potential to be optimized using GenAI. At the same time, prior industrial revolutions have taught that easing work through tool usage could result in job displacement and job loss—which is certainly not the outcome that would convince knowledge workers to start using these tools. Hence, understanding their perspective is pivotal to maximize the collective and organizational benefit from implementing GenAI at their workplaces. The first part of the dissertation comprises two studies focusing on special circumstances of the workplace environment and general influencing factors on GenAI acceptance and usage at the workplace. The first study investigated implications of the recent COVID-19 pandemic on the digital transformation in organizations, a strong driver for the implementation of new technologies like GenAI. Drawing on results from interviews with 15 knowledge workers, the first study revealed that the impact was mainly associated with remote and hybrid work settings. Through the sudden need of digitalization, employees and leaders upskilled in digital working, developed digital formats, and became more susceptible to new technologies. The changes in work setup, personal capabilities, individual needs, and strategic thinking set the perfect stage for new technologies like GenAI to flourish soon after. The second study is a survey study on drivers and barriers of GenAI tool usage at the knowledge worker workplace from an individual perspective. Based on results from previous studies on GenAI tool usage, influencing factors were identified that proved to have a significant effect on users’ behavioral intention to use and actual system usage. With these influencing factors, the established technology acceptance model “Unified Theory of Technology Acceptance and Usage” (UTAUT) was extended with seven additional constructs and employed in the development of a questionnaire for the survey. Data retrieved from 435 valid survey responses from active knowledge workers was analyzed to assess the measurement model and the structural model. Using partial least squares structural equation modelling (PLS-SEM), it was identified that the most influential factor for user intention is performance expectancy, i.e., benefit for users from using GenAI tools. Surprisingly, the study revealed that trust has an ambivalent effect. While trust in GenAI was mentioned as an important factor for using GenAI tools in comments, the hypothesized effect of trust on behavioral intention to use was significantly negative. A potential explanation was found in the fear of AI overreliance when GenAI tools become more trustable. Implied by strong natural language capabilities of GenAI, most GenAI tools are presented as conversational agents (CAs). To examine design and implementation considerations of such CAs at the workplace, the second part of this dissertation comprises three studies. The first study includes a two-part systematic literature research to set the foundation by extracting the state-of-the-art research in CAs for the workplace in terms of application domains and design considerations identified in empirical studies. Based on a total of 45 CA concepts for the workplace, design considerations were extracted and consolidated. Using the Design Science Research methodology, the second study developed and validated a prototype of a CA for workplace learning. Based on semi-structured interviews informed by the Decomposed Theory of Planned Behavior, participants confirmed previously identified design considerations and provided additional insights regarding design and implementation. To conclude this part, the third study zooms into the topic of CA resilience through intentional breakdown handling. Considering strategies employed in research and productive artifacts, a taxonomy was developed to organize and intentionally design breakdown handling strategies. In summary, this dissertation identifies a variety of influencing factors for the acceptance and usage of generative and conversational AI at the workplace. Combining considerations from the organizational and individual perspective, the findings advance the knowledge on AI usage at the workplace, human-AI interaction, and AI system design. Moreover, practical insights help to sharpen the focus on drivers and mitigate barriers to effectively and successfully implement GenAI at the workplace.