On the other hand, chatbots never get tired, are always equally friendly, and can be very knowledgeable if designed accordingly (dimension 1). They also have been shown to be rather charming and persuasive. A recent study comparing chatbots with doctors, for example, found that the chatbots were not only more competent in their diagnosis, but also were perceived as being more empathetic by the patients (Ayers et al. 2023). As the organization chooses between the following workflows, it should evaluate the most promising process designs alongside their business objectives. Workflow Configurations: Designing Processes with “Humans in the Loop” Be it for quality reasons or other business objectives outlined above, fully automated chatbots with absolutely no human oversight are rare and probably are nothing to be desired. For this reason, we now turn to our six workflow configurations that specify how human operators collaborate with GenAI technology. Our six configurations might remind the reader of the levels of autonomous driving. However, while autonomous driving levels are increasing in the responsibility the AI has in driving the car (from level 0: fully human to level 5: fully AI-based) all of our six configurations rely on an interplay between humans and AI. What changes is the division of labor between the two: in configuration 1, the human operator performs the work and the AI focuses on quality assurance and feedback while in configuration 6 it is exactly the opposite. Figure 1 shows the responsibility distribution for response creation and response evaluation performed by the human and agent for each configuration. We illustrate the six configurations for the hypothetical scenario of a patient seeking help in preparing for an upcoming surgery. In such a scenario, the patient may want to know what and until when she is allowed to eat, when she should arrive at the hospital, and when she can expect to go home and resume her normal activities. Figure 1: Responsibility distribution for response creation and response evaluation between human and AI