Aelix Echo · UX Design
Interfaces your operators
actually open every morning.
Design systems, React components, and interfaces designed to WCAG 2.2 AA for enterprise AI products, designed to WCAG 2.2 AA from the first screen.
Agentic AI · Design agents · Now live
Drafts, critiques, audits. The designer commits every change.
Design judgment is not an autonomous task. Aelix Design agents do the surrounding work: drafting variants, auditing accessibility, synthesizing research, surfacing pattern reuse opportunities from your existing system. Every output is a suggestion the designer reviews. The agent never commits a change to your design system, your tokens, or your component library.
Design Critique Agent.
Reviews a draft against your design system tokens, spacing scale, contrast rules, and component conventions. Surfaces inconsistencies and cites the specific token or rule each one breaks.
Accessibility Audit Agent.
Runs the WCAG checklist over a draft. Flags contrast failures, focus traps, missing labels, and keyboard issues. Every flag cites the specific element and the rule. Suggests a remediation. The designer applies it.
Research Synthesis Agent.
Reads through usability sessions, support tickets, and product analytics. Clusters themes and cites the specific quotes or sessions behind each theme. Produces a defensible synthesis, not a manufactured insight.
Pattern Reuse Agent.
Watches for new screens that re-implement an existing pattern. Surfaces the canonical component in your system and the small adjustments that would make it fit. Pure suggestion. The designer decides.
Trust architecture
Assistive only.
Every agent output is a suggestion in a draft state. Nothing commits to your design system, your repo, or your published library without a designer's explicit action.
Grounded in your system, not the model's memory.
The agent reads your tokens, your component library, your research corpus, your accessibility ruleset. It never invents a guideline that does not exist in your system.
Citation on every claim.
Every critique cites the design token or rule. Every research theme cites the source session. Every accessibility flag cites the WCAG criterion.
Iteration without surprise.
Agents are scoped per project. Their drafts and suggestions live next to your work, never on top of it. Your existing design review, design tokens, and accessibility process remain the source of truth.
How the agents behave
Critique and Accessibility Audit agents work as inline suggestions in your design tool. Research Synthesis and Pattern Reuse keep the same assistive posture. We do not ship agents that mutate your design system, your tokens, or your component library. That is the designer's role.
The thesis
A model is only useful when a person can find it, understand it, and trust what it says.
Most AI products fail in the interface, not the model. Our work begins where the prompt ends, turning capability into something a team can adopt, defend, and rely on in the work that actually pays.
What we make
User research that survives contact with engineering.
We sit with the people who'll actually use the product. We map the work, the language, and the constraints. The research outputs become decisions, not slides.
Interaction patterns built for models, not buttons.
Prompting, confidence display, error recovery, human handoff. The patterns most teams reinvent badly, we ship as a library, refined across projects.
Design systems that scale to the org, not just the team.
Token-driven. Code-paired. Versioned like software. Your brand stays consistent across every surface your AI touches, including the ones built by other teams.
Accessibility as the default, not a final audit.
WCAG 2.2 AA built into every flow. Keyboard parity, screen reader QA, color and motion preferences, all handled before the design ships, not retrofitted after a complaint.
Production-ready, end of story.
Figma files, design tokens, accessible React components, Storybook docs, and the conversation with your engineering team. We ship what your team needs to ship.
Patterns we ship
Six interaction patterns most teams reinvent badly, shipped as a library.
A glimpse of the work
Risk review
Risk review surface · enterprise AI workspace
What you get
Researched
Designed with the operators who use the tool
Accessible
WCAG 2.2 AA from the first screen
Systemized
Tokens and components your engineers can ship
How we work
A short engagement, a slow handoff, and a long quiet.
Discover
We sit with operators, map the work, and bring back what the model needs to prove out.
Design
Patterns, flows, and a working design system: drafted in Figma, validated in prototype, decided on the floor.
Build
Accessible React components, design tokens, and Storybook docs handed to your engineering team, plus the working conversation that makes the handoff land.
Adopt
Rollout, telemetry, and the iteration that turns a launch into a habit. We stay through the part that matters most.
A note from the practice
We don't ship demos. We ship work that the operations team takes home on a Tuesday and forgets they're using by Friday, because by Friday it's just the way the work happens.
The Aelix Echo design practice
When the prototype works but the team won't open it, we're here.
45-minute working session · No deck, no pitch