AI-driven onboarding: NovaNotes' context-aware coach increased activation by 34%

AI · 6 min read

NovaNotes, a seed-funded note-taking app, noticed many users never completed their first structured document despite spending time in the editor. The product team hypothesized users needed contextual scaffolding instead of a static tour, so they built a lightweight assistant that suggests next steps based on document type and user activity patterns.

The assistant uses a small retrieval-augmented generation (RAG) pipeline running on the startup's private embeddings index to suggest headings, templates, and example language relevant to the document. To avoid overreach, designers added granular controls: an explicit 'help' affordance, rate limits, and inline explainers showing why a suggestion appeared. Iterative A/B tests measured both activation and long-term retention.

Results showed a 34% lift in 7-day activation and a neutral effect on 30-day churn, indicating the coach improved first-session outcomes without becoming a crutch. The team documented implementation costs, latency impacts, and created a playbook for safe AI UX that other startups can copy: minimal surface area, explainability, and user opt-outs.