Design AI-first interfaces that build ongoing relationships through memory, trust evolution, and collaborative planning, not just isolated screen interactions
Agentic UX Design (Relationship-Centric Interfaces) is a design-guidance skill for building AI-first products that cultivate ongoing relationships rather than optimizing isolated screens. It addresses the problem of session-amnesiac software: systems that make users "start over" every session instead of learning, remembering, and evolving alongside them. The skill articulates a paradigm shift from screen-centric to relationship-centric design and is intended to be used only when explicitly requested for agentic or relationship-centric UX work, not general UI tasks.
The material is organized around five pillars—a memory revolution (from static preferences to behavioral, emotional, and temporal context), trust as a design material (a three-stage evolution from transparency to selective disclosure to autonomous action), relationship-centric architecture, collaborative planning, and measuring relationship quality. Supporting files (REFERENCE, CHECKLIST, EXAMPLES) add technical patterns with TypeScript illustrations such as behavioral event streaming, a contextual memory graph, tiered (hot/warm/cold) memory loading, and privacy-preserving memory with user controls, differential-privacy noise, and PII scrubbing. It cites external research to motivate its design principles and includes explicit guidance on when not to apply the approach (one-off transactions, static content, privacy-sensitive or consistency-critical contexts).
Target users are product designers, UX researchers, and engineers building AI applications, chatbots, or agent systems with repeated long-term interactions who want to design for memory, trust, and co-creation. It is conceptual and reference material illustrated with example code; it does not execute anything and carries no operational risk.
Only when a user explicitly asks for agentic UX, relationship-centric design, or designing for memory, trust evolution, or long-term relationships—not for general UI/UX tasks.
Memory revolution (contextual intelligence), trust as a design material (transparency to autonomy), relationship-centric architecture, collaborative planning, and measuring relationship quality over conversion metrics.
Yes—reference files provide TypeScript patterns for behavioral event streaming, a contextual memory graph, tiered memory loading, and privacy-preserving memory, but as illustrative examples rather than a runnable library.
It emphasizes user control over what is remembered or forgotten, retention policies, memory export, differential-privacy noise for pattern learning, and automatic PII scrubbing.
For simple one-time transactions without accounts, static content sites, systems where memory raises privacy concerns, or interfaces where consistency matters more than adaptation (e.g., medical equipment).
Quick Setup:
.claude/skills/Repository
bencium/bencium-marketplace