AI Assistant UX: How Query-First Design Improved Response Relevance in Draftly

AI · 6 min read

Draftly, an AI-native writing tool, faced complaints about irrelevant or invented content when users issued broad prompts. Rather than only tuning model parameters, the product team introduced a query-first interaction: the assistant now asks structured clarifying questions when prompts lack key constraints. The design pattern reframes model uncertainty as a collaborative clarification step.

The redesign includes an inline clarifier that detects ambiguous prompts and surfaces three suggested clarifying questions (audience, tone, constraints) while allowing users to skip or edit them. Designers balanced latency and flow by making clarifications optional and showing a preview of the expected output length and style. The UI uses subtle animations to keep the assistant visually present without interrupting the user's typing rhythm.

Metrics after rollout show a 24% reduction in content revision cycles and a 29% increase in user-reported relevance scores on surveys. Designers report increased trust and longer session times, but note costs: slightly increased average time-to-output and additional classification logic. Draftly’s approach highlights a core UX principle for AI assistants — use lightweight interaction patterns to resolve model ambiguity rather than expecting raw model improvements to fix interface-level problems.