What UX principles do workplace users want in AI agents?

The paper proposes a design framework of eight core UX principles for human-AI agent interaction in the workplace, each paired with "underlying criteria" that it presents as "actionable guardrails for designers and software engineers." The discussion reports that "business users prioritize principles centered on human control, reliability, context-awareness, and safety," and calls these "practical necessities that directly shape user acceptance and effective collaboration with AI agents" rather than "theoretical ideals."

The reasoning starts from a gap in existing practice. Traditional UX methods were "developed for non-agentic systems producing predictable, human-controlled outcomes," and the introduction argues they cannot account for agents that define objectives, execute multi-step plans, make decisions, and take over "organizational roles previously held exclusively by humans." The framework is meant to help enterprises integrate agents "while maintaining meaningful human oversight, fostering user trust, and adhering to organizational governance requirements." To derive it, the authors combine a participatory design workshop, paper-and-pencil exercises, expert review, meta-analysis, and in-depth interviews, and they describe the result as identified and validated through that combination.

Against the nearest notes, this adds a user-side and designer-side view to arguments made mostly from agent behavior. Should AI systems stay collaborative rather than fully autonomous? argues architecturally that humans should stay in the loop; here the same concerns appear as what workplace users say they want, and as criteria a designer can apply. The stress on human control fits the tension in Does machine agency exist on a spectrum rather than binary?, where users welcome convenience but resist ceding decisions, although this paper does not place its principles on that spectrum. The reliability priority sits beside the measured failures in Why do AI agents fail at workplace social interaction?, but the two are different kinds of evidence: one measures what agents can do, the other what users prioritize.

The excerpt leaves most of the framework's substance out. It does not name the eight principles or the underlying criteria, and it names only four themes that users prioritize. It gives no participant counts, no roles or organizations, and no account of how prioritization was measured or which method produced it. The claim that these principles "directly shape user acceptance" is asserted in the discussion, and no adoption or behavioral outcome appears in the excerpt. The introduction cites controllability, explainability, ethics, privacy, transparency, and collaboration from prior literature, and the excerpt does not say how the eight principles map onto that list. At this strength, the paper supports one narrow reading: in this study's workplace setting, stated user priorities cluster around oversight and dependability, and the authors offer a structured checklist for designing to them. It does not show that following the checklist improves trust or adoption.