Now and even more in the future humans will be interacting with robots and virtual digital agents that have conversational capabilities. For a long time, these systems where only developed in experimental settings. But now they are increasingly employed in real life settings (Pitsch, 2016, 2019). In this presentation, I will show some of the issues of interacting with virtual agents that arise when they are employed in the wild. Based on the methodology of video ethnography and multimodal conversation analysis (Goodwin, 2013; Mondada, 2014; B. Due, 2017; B. L. Due et al., 2019), and video data from different research projects concerned with smart glasses, humanoid robots, AI apps and smart home systems, the presentation discuss how far these systems are from being relevant, ratified, conversational agents.