What vocabulary do AI UX designers actually use most?

UX Design Institute's glossary, authored by Emily Stevens, sets out to name "100 terms all designers should know" for AI UX; read together, its terms are not a flat list but fall into three recurring concerns. One cluster names autonomous, goal-directed behavior — Agentic AI ("AI systems that can autonomously plan actions and execute tasks to achieve a goal"), Autonomous agents, Goal-oriented AI systems, and Intent-based interaction, which "allows users to communicate what they want to achieve without specifying exactly how the system should do it." A second cluster names the mechanisms built to govern that agency: Guardrails ("rules or constraints built into AI systems to guide their behaviour and prevent harmful or inappropriate outputs"), Human-in-the-loop AI, Human override controls, AI trust signals, and Trust calibration, defined simply as "designing AI systems so that users trust them at the right level." A third cluster names the conversational/natural-language interface as the default surface for all of this — Conversational AI, Conversational UI, Conversational workflow, Copilot interface, LLM interface design, Natural language interface — with AI experience design (AIX) named as the umbrella discipline covering all three, "an emerging discipline focused on designing user experiences for AI-powered products and features" that, unlike traditional UX, "also considers factors unique to AI systems, such as transparency, trust, automation, uncertainty and human-AI collaboration."

The glossary's own reasoning is definitional rather than argumentative — it does not claim these three clusters are the taxonomy, it simply accumulates terms practitioners are using, and the clustering is read off which concepts keep recurring across its entries. That the same three concerns (agency, oversight, conversation) surface independently across a vocabulary list compiled for working designers is itself the pattern worth noting: it suggests the terms practitioners reach for to talk about AI UX are organized around the same fault lines that academic and industry frameworks have converged on separately.

That convergence is visible against How should AI systems be designed across different interaction phases?, which groups guidance by interaction phase rather than by concern, but covers much the same ground — guardrails and override controls map onto Amershi's when wrong phase, trust signals and calibration onto the ongoing trust-building that runs across all four phases. It also lines up with What UX principles do workplace users want in AI agents?, where human control and safety are exactly the concerns the glossary's oversight cluster names, and with How should AI interfaces handle the shift from doing to supervising?, whose intent surface is the same intent-based interaction the glossary defines. Where this glossary adds something distinct is vocabulary coverage breadth: it is the only one of these sources naming the conversational-interface cluster (Conversational UI, LLM interface design, Copilot interface) as its own concern alongside agency and oversight, rather than folding interface modality into the agency or control discussion.

What the excerpt does not establish is which of these terms are stable industry consensus versus one author's synthesis, or whether practitioners outside this glossary use these terms consistently; a glossary entry is a proposed definition, not evidence of adoption. It also gives no guidance on how the three clusters trade off against each other in a concrete design — for instance, whether more guardrails necessarily cost intent-based flexibility. The implication is narrow: this is useful as a shared-vocabulary anchor when writing about AI UX for a mixed audience, not as evidence that these are the right or only three concerns.