Agentic AI systems—ensembles of data-processing, machine learning and generative components that engage in autonomous decision-making—are rapidly moving from the lab into real-world practice. Yet most applications remain “socially naïve,” overlooking power relations, norms and situated work practices. This workshop will gather researchers and practitioners to: (1) explore how agentic systems are being interpolated into social and organizational contexts; (2) identify social-theoretic lenses (e.g., third-wave HCI, human-AI teaming, post-/more-than- human design) that can inform new agentic architectures; (3) seed collaborations that combine software engineering with humanities and social-science insights. Given the focus of CASCON 2025 on the modern computing landscape, we see this work as relevant to the collaborative development of novel design patterns, testing regimes and governance methods that support the responsible deployment of robust, context-aware agentic applications.