Patterns

Progressive Disclosure

The AI design pattern library. One citation-backed pattern, read in 90 seconds.

A pattern is one researched, cited guideline for designing AI, read in 90 seconds.

Category

Track

29
Design streaming AI responses for screen readers, not just for the eye
Accessibility
accessibilityscreen-readersstreamingaria-livechatbotsconversational-ui
Track A
28
Don't outsource WCAG compliance to an AI overlay
Accessibility
accessibilitywcagoverlayscomplianceftctrust
Track A
27
Treat AI generated alt text as a draft, not a deliverable
Accessibility
accessibilityalt-textscreen-readersprovenancetrustimage-description
Track A
26
Accessibility requests are legal now: the European Accessibility Act covers your AI features
Accessibility
accessibilityEAAcompliancechatbotsWCAGregulation
Track A
25
Design the handoff to a human before the chatbot needs it
Error recovery
conversation designchatbotsescalationerror recoverycustomer experience
Track A
24
Let the model cluster the transcript, keep the human naming the themes
Research methods
ux researchqualitative analysisthematic analysisllmresearch ops
Track A
23
Design remote studies to catch chatbot generated participant answers
Research methods
ux researchsurveysdata qualityai detectionstudy design
Track A
22
Use synthetic users to prepare research, not to replace it
Research methods
ux researchsynthetic usersevidencesycophancystudy design
Track A
21
Evals are the new usability test: design teams define what good looks like, then test for it
Tools and integrations
evalsai-agentstestingdesign-opsdesign-tools
Track B
20
Shipping AI prototypes with v0 and Vercel: set access before you share the link
Tools and integrations
v0vercelprototypingdeploymentdesign-tools
Track B
19
Cursor for design engineers: carrying design intent into code
Tools and integrations
cursordesign-systemsmcpdesign-engineering
Track B
18
Using Claude artifacts in a design workflow: what fits and what doesn't
Tools and integrations
claudeartifactsprototypingdesign-tools
Track B
17
Prompt-to-prototype tools: when Lovable-class tools beat hand design
Tools and integrations
prototypinglovablefidelitydesign-tools
Track B
16
AI prototyping in Figma: where it helps and where it lies
Tools and integrations
figmafirst-draftfigma-makeprototyping
Track B
15
What MCP changes about designing AI integrations
Tools and integrations
mcpconnectorspermissionstrust
Track B
14
Give users memory and personalization controls, not surprises
Personalization
memorypersonalizationuser control
Track A
13
Don't present chain-of-thought as an explanation
Explainability
explainabilitychain-of-thoughttransparency
Track A
12
Design the wait as part of the interaction
System status
waitingprogress-indicatorsfeedback
Track A
11
Make every AI action reversible or checkpointed
User control
reversibilityundoagents
Track A
10
Design for appropriate reliance, not maximum trust
Calibrated trust
calibrated-trustoverreliancetrust-in-automation
Track A
09
Don't dress the AI up as a person
Honest representation
anthropomorphismtransparencytrust
Track A
08
Layer AI explanations: short in the flow, deep on request
Explainability
explainabilitytransparencycitations
Track A
07
Help users specify intent without prompt gymnastics
Intent specification
intent specificationprompt controlsarticulation barrier
Track A
06
Show the plan before the agent runs it
Agent oversight
agentsoversightprogress feedback
Track A
05
Ask a clarifying question instead of guessing, but only when it matters
Mixed initiative
mixed initiativeclarifying questionsuncertainty
Track A
04
Treat AI errors as the default case, not the edge case
Graceful failure
graceful failureerror handlingrecovery
Track A
03
Set capability expectations before the first prompt
Expectation setting
expectation settingonboardingcalibrated trust
Track A
02
Design for honest pushback, not agreement
Calibrated trust
calibrated trustsycophancypushback
Track A
01
Audit AI features against the classic heuristics and the AI guidelines
Usability heuristics
foundationsheuristicsexpectation setting
Track A

Progressive Disclosure

Design streaming AI responses for screen readers, not just for the eye

Don't outsource WCAG compliance to an AI overlay

Treat AI generated alt text as a draft, not a deliverable

Accessibility requests are legal now: the European Accessibility Act covers your AI features

Design the handoff to a human before the chatbot needs it

Let the model cluster the transcript, keep the human naming the themes

Design remote studies to catch chatbot generated participant answers

Use synthetic users to prepare research, not to replace it

Evals are the new usability test: design teams define what good looks like, then test for it

Shipping AI prototypes with v0 and Vercel: set access before you share the link

Cursor for design engineers: carrying design intent into code

Using Claude artifacts in a design workflow: what fits and what doesn't

Prompt-to-prototype tools: when Lovable-class tools beat hand design

AI prototyping in Figma: where it helps and where it lies

What MCP changes about designing AI integrations

Give users memory and personalization controls, not surprises

Don't present chain-of-thought as an explanation

Design the wait as part of the interaction

Make every AI action reversible or checkpointed

Design for appropriate reliance, not maximum trust

Don't dress the AI up as a person

Layer AI explanations: short in the flow, deep on request

Help users specify intent without prompt gymnastics

Show the plan before the agent runs it

Ask a clarifying question instead of guessing, but only when it matters

Treat AI errors as the default case, not the edge case

Set capability expectations before the first prompt

Design for honest pushback, not agreement

Audit AI features against the classic heuristics and the AI guidelines