AI UX Patterns: 150+ Interface Design Examples with Demos (AI UX Playground)
Source: AI UX Playground pattern library · undated
Agentic UX: Use when the product can take actions (send, spend, edit, delete) and you need to decide how much authority is safe.
Chat UX: Use when the product is a conversation: composing turns, streaming answers, remembering context, and recovering when a turn fails.
Trust Scaffolding: Use when people must decide whether to believe or rely on the system, from first run through evidence, control, and long-term use.
Citations: Attach verifiable sources, quality signals, and claim previews
Chain of Thought: Reveal step-by-step reasoning behind an answer
Granular Consent: Per-capability permissions with scope, expiry, revoke, and drift
Failure Disclosure: Say clearly when the system cannot answer or a tool failed
Data Ownership & Control: User control over AI data usage
Source Browser: Inspect retrieved sources and context beside the answer
Streaming: Show replies token-by-token as they generate
Response Refinement: Modify AI responses with contextual actions
Chat Artifacts: Open generated docs and code in a side panel
Follow-up Chips: Suggested next turns
Interrupt and Resume: Stop mid-response and continue with context
Memory Scope Toggle: Set memory persistence per message
Tool Use: Show when the model uses tools and what came back
Autonomy Budgets: Time- or action-bounded unattended runs
Human in the loop: Require human approval before AI acts
Suggest / Confirm / Execute: Three autonomy modes for AI agents
Per-Action Autonomy: Autonomy scoped per capability, not per app
Feedback: Collect thumbs and comments to improve answers
Conversation Summary: Auto-summarize long chats
Variation Picker: Pick among side-by-side generated options
Smart Diff: Highlight and review what the model changed
Voice Commands: Trigger actions via voice commands
Interruptibility: Single gesture to pause or cancel
Cost Transparency: Show the price of an AI operation before or as it runs
Running Meters: Live token, cost, and quota counters during execution
Processing Time Estimates: Expected wait times
Hard Budget Ceilings: Enforceable spend caps across runs and sessions
Lines of inquiry this paper opens 24
Research framings built by reading the notes related to this paper — the questions it feeds into.
How do confidence signals and expertise language mislead users' trust calibration?- How do stochastic outputs create mental model and trust problems for users?
- How does an action receipt differ from capability disclosure and explainability in the trust workflow?
- How do gradients compare to mascot characters in user trust studies?
- Can ambient visual signals replace explicit negotiation surfaces for AI trust?
- How do users conflate intuitive trust with reasoned evaluation of AI outputs?
- Which AI mistakes trigger the strongest trust withdrawal in users?
- How does anthropomorphism affect user trust in AI systems?
- Does showing confidence levels make users trust agents less accurately?
- How does vendor-prescribed design doctrine compare to frameworks derived from user research?
- How do interface patterns compare to organizational AI oversight processes?
- Can interface design nudge users toward verification of AI responses?
- How do pre-action UX patterns prepare users for staged autonomy expansion?
- Which design choices boost engagement but increase user vulnerability?
- How do Amershi's interaction-phase guidelines compare to concern-based AI UX frameworks?
- How can designers make control feel available without constant user oversight?
- Does moving users through autonomy levels require different interface patterns at each stage?
- How often do products ship all three control patterns together in practice?
- How do autonomy budgets map onto OpenAI's Model Spec scope requirements?
- What evidence shows sliders outperform toggles for setting agent independence?
- Do users actually distinguish per-action autonomy scoping from per-app scoping?
- Should a product override user autonomy settings when agent confidence is low?