Conversational UI
What is Conversational UI?
Conversational UI is a design pattern where users interact with AI through natural language (text or voice) instead of traditional menus, forms, and buttons. Rather than learning a product's interface, users simply describe what they need and the AI interprets their intent. Conversational interfaces come in many forms: text-based chatbots like ChatGPT and Claude, voice assistants like Siri and Alexa, embedded AI assistants like Slack AI and Microsoft Copilot, and hybrid interfaces that combine chat with traditional UI elements.
Problem
Traditional graphical interfaces require users to learn specific navigation patterns, menu hierarchies, and form layouts. As AI products grow more capable, the gap between what the system can do and what users can discover widens. Users prefer asking for what they need in plain language, but poorly designed conversational interfaces frustrate them with robotic responses, lost context, and dead-end conversations.
Solution
Design conversational interfaces that understand natural language, maintain context across multiple turns, and respond in a natural, human-like way. Support both text and voice input where appropriate.
Real-World Conversational UI Examples
Implementation
Practice in Courses
When to use Conversational UI, and when it backfires
Use it when
- The task is open-ended or hard to express through a form or menu: the user doesn't know the exact term, or the input space is too large to enumerate (support, search, generation).
- Intent genuinely varies turn to turn and benefits from follow-up, clarification, and memory of what was just said.
- Plain language is actually faster than the UI alternative, e.g. 'find the invoice from Acme last March' beats five filters.
Don't, or minimize, when
- The task is a known, finite set of choices. A button, toggle, or form is faster and less error-prone than parsing a sentence. A chat box wrapped around a 3-option workflow is worse UX, not better.
- The action is high-frequency and precise (changing a setting, picking a date). Typing a request is slower than a direct control.
- The model is slow or unreliable for this task. Conversation amplifies latency and errors; a deterministic UI degrades more gracefully than a chat that stalls or hallucinates.
The trap
The blank chat box. A bare text input with no guidance looks magical in a demo and strands real users who have no idea what to type or what the system can do. Its cousin is the over-conversational UI, where core features get buried behind 'just ask' and become undiscoverable. A conversation is an input method, not an excuse to delete the interface.
Take it into your own product
- 1
Never ship a blank chat box.
The empty state is the whole game. Suggested prompts turn 'what can this even do?' into a single click, and they teach users the system's range without a manual. A blank input is not minimalism, it's an unfinished feature.
- 2
Show the work: typing, thinking, streaming.
Silence reads as broken. Streamed tokens and honest status cues (searching, generating) make latency tolerable and the system feel alive. A frozen spinner is the fastest way to make a capable model feel dead.
- 3
Mix chat with buttons and cards. Don't force everything into text.
When the next step is a finite choice, render a button, not a sentence the user has to compose and the model has to parse. The best conversational UIs let people click or type at every turn, whichever is faster.
- 4
Design the misunderstanding, not just the happy path.
A specific clarifying question beats a generic error every time. 'Did you mean X or Y?' keeps the user moving; 'I didn't understand that' sends them away. The recovery turn is where trust is won or lost.
- 5
Always give an exit.
When the AI hits its limit, hand off to a human or a structured flow with the context preserved. A conversation the user can't escape, and has to restart by repeating themselves, is worse than the form you replaced.
Save Conversational UI as a Claude skill
Saved skills collect on your dashboard, ready to download one at a time or as a pack for your repo. Once a skill is in place, Claude Code applies it whenever you work on a surface this pattern covers.
Check if your product already has this pattern
Upload a screenshot. We'll tell you which of the 38 patterns your AI interface uses and where the gaps are.
Audit My DesignTake all 38 patterns as Claude skills
Get the whole library as skill files, so Claude applies these patterns while it builds instead of after you catch them in review.
- One skill file per pattern, all 38
- Drop them in .claude/skills and Claude applies them as it builds
- Daily AI UX newsletter (unsubscribe anytime)
More in Natural Interaction
Progressive Disclosure
Gradually reveal information, options, or AI features to reduce cognitive load and simplify complex tasks.
Multimodal Interaction
Combine voice, touch, gesture, text, and visual input for natural interaction.
Context Switching
Smooth transitions between tasks or topics while maintaining conversation continuity.
