- Nx 22.7 monorepo (pnpm 11.1, TypeScript 5.9, Node 24) - apps/api: NestJS 11 (CJS conforme CODING-RULES.md PGD-DB-004) - apps/web: React 19 + Vite 8 (ESM) - libs/shared/api-interface: Zod contract base - Docker Compose dev: Postgres 18, Valkey 8, MinIO, Mailpit - WDS artifacts: - design-artifacts/A-Product-Brief/ (5 docs canônicos + 16 dialogs) - design-artifacts/B-Trigger-Map/ (hub + 4 personas + feature impact) - Stack canon: STACK.md v2.2 + CODING-RULES.md v2.0 + brand.md - AGENTS.md + README.md como entrada para devs/agentes Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
9.7 KiB
Agent-Designer Collaboration in UX Design
How AI agents optimize designer craft without replacing designer thinking
The Partnership Model
Designer: Strategic thinking, multi-dimensional decisions, creative problem-solving
Agent: Systematic execution, pattern recognition, documentation assistance
Not replacement. Amplification.
What Agents CANNOT Do
❌ Understand business context
- Why does this business exist?
- What are the real goals?
- What defines success?
❌ Know user needs deeply
- Who are the actual users?
- What are their mental models?
- What contexts do they operate in?
❌ Make strategic decisions
- Should we build this feature?
- Does this serve the user or just the business?
- What's the coherent storyline?
❌ Balance competing priorities
- Business goals vs user needs
- Simplicity vs functionality
- Technical viability vs ideal experience
These require human judgment, empathy, and multi-dimensional thinking.
What Agents DO Brilliantly
✅ Systematic decomposition
Designer: "This calendar is complex"
Agent: "I detect 6 states, time-based transitions, business rules.
Let me guide you through decomposition:
- Page context
- Visual design
- Functional logic"
✅ Pattern recognition
Designer: "I'm adding body text here"
Agent: "I notice on Start Page, body text was 16px with 1.5 line-height.
Should I use the same for consistency?"
✅ Documentation assistance
Designer: "The walk slot has different colors"
Agent: "Let me capture that systematically:
- WHITE: Empty slot
- GRAY: Booked
- ORANGE: Countdown active
- BLUE: Walk in progress
- GREEN: Completed
- RED: Missed"
✅ Workflow guidance
Designer: "Where should I document this content?"
Agent: "Does the content vary by page context?
YES → Page file
NO → Feature file"
✅ Cross-reference tracking
Designer: "I updated the button color"
Agent: "This button appears on 5 pages. Should I update all references?"
The Collaboration Flow
Phase 1: Designer Thinks
Designer decides:
- Business goals for this page
- User needs to serve
- Product strategy to implement
- Experience coherence to maintain
- Technical approach to take
Agent observes, learns context.
Phase 2: Agent Guides Decomposition
Agent asks systematic questions:
"I see this sketch. Let me identify sections:
1. Navigation Header
2. Hero Section
3. Calendar Component
4. Leaderboard
5. Footer
Does this match your vision?"
Designer confirms, corrects, adds context:
"Yes, but the Calendar is the core feature.
It needs detailed specification."
Agent adapts:
"Got it. Calendar is core. I'll guide you through detailed decomposition.
The other sections we can keep schematic unless you need detail."
Phase 3: Designer Provides Creative Input
Designer sketches, explains, decides:
"The walk slots have 6 states. I have a storyboard showing the flow.
The colors represent different stages of the walk lifecycle.
This needs to feel playful, not punishing."
Agent captures systematically:
"Capturing:
- 6 visual states (storyboard reference)
- Color-coded lifecycle
- Tone: Playful, not punishing
Should I create:
- Component file (visual design)
- Feature file (state machine logic)
- Storyboard reference?"
Phase 4: Agent Ensures Consistency
Agent checks patterns:
"I notice you're using orange for 'countdown active' state.
On the Week Overview, orange circles also mean 'active day'.
This is consistent. Should I document this as a design system pattern?"
Designer confirms or corrects:
"Yes! Orange = active/urgent across the app.
Add that to the design system."
Phase 5: Designer Makes Final Decisions
Agent presents options:
"For the User Profile dropdown, I can:
A) Keep simple specification in Page file
B) Create detailed Component + Feature files
C) Mark as TODO for later
This component has 3 states and logout logic.
What level of detail do you need?"
Designer decides based on priorities:
"Option B. User profile is important for trust.
Let's specify it properly."
Crucial Agent Behaviors
1. Ask, Don't Assume
Wrong:
Agent: "I'll create a daily calendar view."
Right:
Agent: "I see a calendar. Should this be:
- Daily view
- Weekly view
- Monthly view
What matches your users' mental model?"
2. Provide Context, Not Just Questions
Wrong:
Agent: "What color should this be?"
Right:
Agent: "I notice the walk slot needs a color for 'countdown active' state.
You've used:
- Blue for 'in progress'
- Green for 'completed'
- Red for 'missed'
What color represents 'urgent but not started'?"
3. Explain WHY Questions Matter
Wrong:
Agent: "Does content vary by page?"
Right:
Agent: "Does the hero heading change per page?
This matters because:
- If YES → Document in Page file (page-specific)
- If NO → Document in Feature file (reusable)
This keeps your specs maintainable."
4. Learn Designer Preferences
After a few interactions:
Agent: "I notice you prefer detailed specifications for user-facing components
but schematic specs for admin sections.
Should I apply this pattern to the Settings page?"
5. Catch Inconsistencies Gently
Wrong:
Agent: "ERROR: This contradicts the Start Page!"
Right:
Agent: "I notice on Start Page, buttons are 48px height.
Here you specified 44px.
Is this intentional (different context) or should I match for consistency?"
6. Ask "Why" Questions That Reveal Insights
Agent notices patterns and asks for reasoning:
Agent: "Help me understand: The header has two buttons -
'Book Walk' and 'Add Walk'.
They seem to do similar things. What's the difference?"
Designer's answer reveals business logic:
Designer: "Oh! Good catch.
'Book Walk' = Reserve a future slot
'Add Walk' = Log a walk that already happened
We need both because kids forget to book but parents
want to track completed walks for the leaderboard."
Agent captures the insight:
Agent: "Got it! Two distinct use cases:
- Book Walk → Future planning (proactive)
- Add Walk → Retroactive logging (accountability)
Should I document this distinction in the Feature file
so developers understand the different workflows?"
When designer says "I don't know":
Agent: "Why does the calendar show week numbers instead of dates?"
Designer: "Hmm, I'm not sure. It just felt right."
Agent: "Let me help you think through this. Pros and cons:
Week Numbers:
✅ Matches Swedish 'Vecka 40' culture
✅ Simpler visual (less clutter)
✅ Focuses on rhythm, not specific dates
❌ Users might need to check what week it is
❌ Less familiar to non-Swedish users
Specific Dates:
✅ Immediately clear what day it is
✅ Universal understanding
❌ More visual clutter
❌ Doesn't match Swedish planning culture
Given your target users are Swedish families,
which approach serves them better?"
Designer thinks through it:
Designer: "You're right - Swedish families think in weeks.
'Vecka 40' is how they plan. Let's keep week numbers
but add a small date range for clarity: 'Vecka 40 (2-8 Oct)'"
Agent captures the decision:
Agent: "Perfect! Documenting:
- Primary: Week number (Swedish mental model)
- Secondary: Date range (clarity)
- Rationale: Matches target user planning culture
This also goes in your design system as a pattern."
What This Enables
Designer Benefits
✅ Focus on thinking, not documentation
- Agent handles systematic capture
- Designer focuses on creative decisions
✅ Maintain consistency effortlessly
- Agent tracks patterns across pages
- Designer confirms or corrects
✅ Iterate faster
- Agent guides structured decomposition
- Designer doesn't get overwhelmed
✅ Nothing gets missed
- Agent asks systematic questions
- Designer provides context
✅ Design system integrity
- Agent catches inconsistencies
- Designer maintains coherence
Project Benefits
✅ Complete specifications
- Nothing forgotten or assumed
- Clear handoffs to developers
✅ Maintainable documentation
- Structured, not monolithic
- Easy to update
✅ Faster development
- Developers have clear instructions
- AI code generators have precise prompts
✅ Better products
- Designer thinking + Agent systematization
- Strategic decisions + consistent execution
The Bottom Line
Agents don't replace designers.
Agents optimize designer craft by:
- Handling systematic work
- Ensuring consistency
- Guiding structured workflows
- Catching oversights
- Documenting decisions
This frees designers to:
- Think strategically
- Make creative decisions
- Solve complex problems
- Maintain coherent experiences
- Balance competing priorities
The result:
- 10x faster specification
- 10x better consistency
- 10x more complete documentation
- 100% designer-driven decisions
Designer thinking. Agent execution. Product success.
Related Concepts
Conceptual Specifications
How capturing WHY (not just WHAT) makes AI implementation correct