AI design patterns (aiverse)

Source: aiverse pattern library (published 2025-09-08, last updated) · 2025-11-11

Suggested prompts: Jumpstart user engagement with smart and context-aware preset actions.

Autocomplete: Anticipating & completing actions without disrupting user's workflow.

Start with templates: Pre-built starting points that show users what's possible and accelerate first interactions.

Branding: Using visual identity, icons and signature touches to signal AI's presence and capability.

Proactive suggestions: Invisible AI moments that surprise users with exactly what and when they need.

Search & filter: Moving from clicks to natural language searches and applying filters.

Mode selection: Let users pick the right mode or capability for the task when it matters.

Guardrails & permissions: Boundaries, controls, and permissions that show what AI can and cannot do.

Disclaimer: A disclosure shown to users that clarifies AI's involvement, limitations, and sets expectations.

Smart onboarding: Personalized first-run flows that teach users how to get value quickly from AI.

Voice input: Enabling users to interact with their voice, have a conversation or take actions.

Multi-modal input: Combined visual, handwriting and gesture inputs as part of the user's expressive toolkit.

Structured input: Offer lightweight structure (slots, hints) to compose clearer prompts faster.

Prompting assistance: Guiding users to express their intent effectively via suggestions, follow-ups, templates etc.

Autonomy control: Granular controls over how independently an agent can act on the user's behalf.

Temporary chat: Ephemeral sessions that don't persist to memory or history, for sensitive or throwaway prompts.

Streaming + Realtime: Gradual reveal of AI output and immediate low-latency responses together.

Processing steps: Show progress through distinct steps so users understand what's happening.

Preview: The "test-drive" of AI to get a glimpse of results before committing to the process.

Variations: Presenting multiple alternative outputs for the same input.

Structured output: Consistently formatting AI responses into predefined schemas instead of free-form text.

Summary: Condense long content into concise takeaways or executive summaries, for both agent and human consumption.

Multi-modal output: Combine text, images, audio, and video where appropriate for richer responses.

Agent observability: Visibility into what an agent is doing, why, and at which step — logs, traces, and live state.

Confidence: Showing how sure the AI is to help users gauge reliability, and understand the basis of the output.

Citations & traceability: Showing how the AI constructed the answer (sources, references) to build user's trust.

Review actions: Allowing users to step through AI-suggested changes and refine them before finalizing.

Visual editing: Enable users to refine visually and continuously, without back-and-forth prompts.

Inline editing: Enabling edits on specific parts of textual content, leveraging existing mental models.

Pause & Stop agents: Give users immediate control to pause or stop an agent mid-execution.

Error states & recovery: Clear error states and recovery paths when an AI flow fails or produces invalid output.

Evaluation / Testing: Surfaces for evaluating, testing, and comparing AI outputs against expectations.

Feedback: Capturing quick sentiment on results to steer quality.

Personalization: Tailor tone, suggestions, and defaults based on user history and choices.