UX Case Study: Integrating LLM-Based Search into NoteSpark Without Overwhelming Users

AI · 4 min read

NoteSpark's product team wanted to add a semantic search assistant to help users surface insights from large note collections. Early prototypes produced impressive results but also occasional hallucinations and inconsistent phrasing, which risked eroding trust. The design team created a phased integration that balanced generative capabilities with familiar UI patterns.

Key decisions included presenting the LLM assistant as an 'Insight Lens' toggle rather than the default search, showing provenance metadata for generated answers, and offering a one-tap 'source view' that links back to original notes. The team also added conservative guardrails: suggested rewrites were labeled as 'AI suggestions' and a feedback control allowed users to flag incorrect results for retraining.

The staged rollout began with power users and a visible trust dashboard that tracked accuracy feedback. Metrics showed the Lens increased saved-search usage by 30% among early adopters, while overall search satisfaction rose by 12%. The case underscores that integrating AI into UX demands clear signals about capabilities and limitations, as well as easy escape hatches to traditional controls.