Publication
Expecting Automation: How Users Miss Valuable Experiences in Generative AI
Aeneas Stankowski
In: AutomationXP26: Agentic Automation Experiences. Workshop at CHI2026 in Barcelona, 2026.
Abstract
Research on human-AI interaction has documented several phenomena that shape how users engage with generative AI: transactional
interface affordances, cognitive offloading tendencies, over-reliance on AI outputs, and organizational pressures favoring rapid
completion. These are individually well-established, with parallels in the automation complacency and bias literature. They result in a
compounding effect and default orientation toward expecting full automation, or handoff. Drawing on a Research through Design
study of AI-supported corporate goal-setting (N=15) and subsequent informal deployment, we observe that this expectation led
participants to neglect opportunities for agency and steering, even when offered, creating a barrier between experienced value and
chosen engagement. Participants who underwent sustained reflective interaction in structured sessions found it uniquely valuable, yet
in informal use quickly attempted to redirect the system toward generating artifacts. We argue that automation expectation, as a
convergent effect operating prior to engagement, warrants attention as a design problem in automation experience research, and
outline directions for intervention.