The paper reports on an in-situ qualitative study of a persistent, proactive AI agent deployed across multiple teams in a large technology company. The agent was designed to streamline communication and coordination, but its introduction caused breakdowns in three areas: implicit collaborative workflows, relational norms for the non-human actor, and the redistribution of trust and agency.
These findings are interesting because they move beyond simulated lab studies to capture real-world frictions that arise when autonomous agents enter human workspaces. The authors use these early signs of negotiation and conflict to propose a research and design agenda focused on preserving human agency in workplaces shared with AI.