How Schedly Rebuilt Trust in Its AI Scheduler Onboarding

AI · 5 min read

Schedly launched with an ambitious assistant that scanned email and calendar to propose meeting times, but the first-week metrics showed a 68% drop-off during onboarding. Users hesitated at permission screens and abandoned when uncertain what the assistant would read or change. Early qualitative interviews flagged two main fears: privacy and unpredictability—users wanted clear guardrails, not opaque automation.

The redesign centered on progressive disclosure and explainability. Rather than a single broad permission screen, Schedly introduced stepwise scopes (read-only calendar preview, suggested text only, automated invites opt-in). Each scope included a concise, example-driven explanation and a “show me what it would do” preview that generated a sample meeting invite from anonymized data. The team also added a privacy toggle that allowed a permissionless demo using synthetic data and a one-tap roll-back control to undo automated invites.

Design and product ran a three-week A/B test with 4,200 new signups. The new onboarding reduced drop-off from 68% to 24% and tripled the percentage of users who enabled automation within their first session. Trust metrics from follow-up surveys rose: perceived clarity of actions increased from 31% to 78%, and 72-hour retention doubled. Key takeaways: show examples, limit initial scope, and give immediate reversibility to mitigate perceived risk.