6: The Agentic AI Imaginary

During the last chapter, we explored how living and interacting with machines had opened up new avenues for engaging with and instructing generative AI applications in and through natural language capacities. We described this emerging conduit for human-machine interaction as ‘ethno-programming’ in and through which senses of social organisation and relationality can be elicited and applied, through generative AI platforms, to various practical tasks, for example, by prompting an LLM to provide life coaching advice or act as a language tutor or editorial assistant in ways that can accomplish an interactive role stance that is bound up with categorially organised and displayed knowledge. Now, the display of such category-bound knowledge may be understood, made sense of, in different ways by human actors as part and parcel of the situated and occasioned engagement and interrogation of generative AI interfaces. This also resonates with more general questions about accuracy, veracity and wider ethical implications that surround the consumption, production and replication of bias (Abid et al., 2021; Afreen et al., 2025; Kotek et al., 2023) within algorithmic and generative AI contexts and beyond, especially when coupled with next-generation social media. Nevertheless, role prompt engineering, a clear example of one strategy that has reflexively and technically bounded sociological features (Housley & Dahl, 2024), points to the developmental possibilities of LLMs to act as both stochastically realised and human-configured companion ‘agents’.