Using Human-in-the-Loop AI to Rescue Onboarding for Tutorly

AI · 4 min read

Tutorly, a marketplace for tutors and students, struggled with onboarding completion—profile setup required nuanced descriptions of teaching style, availability, and sample materials. Fully automated prompts produced boilerplate bios that didn't convert; manual coaching didn't scale. The product team implemented a human-in-the-loop AI flow: tutors drafted bios that an LLM refined into persona-aligned profiles, then a small team of writing coaches reviewed and tweaked high-impact profiles.

The hybrid flow reduced friction: drafts took 90 seconds on average, automated suggestions improved clarity, and coach reviews ensured quality for top-tier listings. Onboarding completion rose 28%, and bookings per new tutor increased 15% in the first two months. Importantly, user trust remained high because edits were transparent and tutors retained ownership over final text.

Tutorly's experiment underscores design principles for sensitive domains: use AI to amplify, not replace, human judgment; preserve user control and transparency; and prioritize high-touch review for entries that materially affect outcomes. The article encourages startups to map which content types need human oversight and to instrument quality feedback loops to continuously tune the AI prompts and guardrails.