Conversational UI
Completing a task as a back-and-forth conversation, with the interface confirming what it understood at each step.
See it
What it is
Conversational UI turns a task into a sequence of turns: the system asks for what it needs, the person answers, and each step confirms what was understood. The best versions are not walls of chat bubbles. They mix plain language with quick replies, date pickers, cards, and summaries whenever structured controls are faster or less ambiguous than typing. Intercom's support bots and Typeform's one-question-at-a-time forms are the same shape wearing different clothes.
Reach for it in support triage, guided setup, booking, and infrequent tasks where people do not know the workflow in advance. A conventional form, table, or command palette is better for comparing many options, editing several fields at once, or repeating a familiar expert task.
The gotcha is hidden state. If the system misunderstands a date or forgets an answer ten messages back, the person needs a visible way to inspect and correct it. Summarize the interpreted details before any side effect, ask only for missing information, preserve the transcript, and never claim an action succeeded until the backend confirms it.
Ask AI for it
Build this booking flow as a conversational UI backed by a finite-state machine with collecting, confirming, submitting, success, and error states. Render messages in a container with role='log' and aria-live='polite'. Ask one question at a time, offer quick-reply buttons for fixed choices, and switch to native date and time inputs when free text would be ambiguous. Keep an editable summary card pinned above the composer, require an explicit Confirm action before submitting, and show Retry without losing answers if the request fails. Never send a backend mutation directly from an unconfirmed free-text message.