UX Collective: Design's alive and kicking. It just got some flashy new names.
Published in June 2026, Nicole Alexandra Michaelis’s piece on UX Collective makes a specific argument against the wave of predictions that AI will hollow out design work: rather than eliminating design roles, the current moment is creating a new class of specialized positions with no direct predecessors in the field.
Michaelis writes from her background as an AI conversation designer, and this shapes the article’s angle. The focus is less on whether designers’ jobs are safe in the abstract and more on what the actual new work looks like for people building products at the intersection of generative AI and user experience.
Five emerging roles
The article identifies five roles she sees forming in product teams:
The embedded AI design consultant studies how an organization’s workforce uses tools, maps conversational and agentic touchpoints, and designs solutions for specific adoption challenges. The work is closer to organizational design than traditional UX—it starts with understanding where AI meets actual workflows rather than where it could theoretically fit.
The agentic UX architect owns the end-to-end logic of how AI agents behave from the user’s perspective: sequencing, fallback states, and how multi-step autonomous tasks feel to the people they serve. This is interface design applied to systems that act over time.
The proactive interaction designer focuses on AI that initiates, anticipates, and suggests rather than waiting for user input—a significant shift from reactive interface design that assumes users know what they want before they ask.
The generative UI system architect builds the design-system infrastructure that allows interfaces to adapt dynamically to context and user state. Rather than rendering fixed components, these systems make real-time decisions about what to show and how.
The trust designer addresses how AI systems communicate uncertainty, explain decisions, and make it possible for users to verify or override outputs. The underlying problem is that AI systems are wrong in ways that are difficult to detect, and the interface needs to make that visible.
Context and limitations
Michaelis frames these roles as a genuine expansion of the discipline’s scope, not a rebrand of existing titles. The argument is structural: agentic AI products generate new user problems that existing role definitions are not equipped to solve.
The article is most useful for design leaders and senior designers thinking through how team structures will change as agentic AI products move from prototype to production. It works better as a map of what’s emerging than as a practical guide to any individual role—the descriptions are brief, and each could support a much longer treatment.