SaaS AI Onboarding
Flow
A scalable, AI-assisted content system that reduced time-to-value and survived a fast-moving product roadmap
The challenge: a product shipping new features every two weeks, a content team of one, and users abandoning before they understood the value.
- Project ContextThe Problem
A SaaS tool with strong feature depth was losing users in the first 7 days - not because the product was bad, but because users couldn't figure out the value fast enough. The onboarding flow was a 12-step checklist that nobody finished. In-app guidance was tooltip text that explained what buttons did, not why users would care.
Secondary problem: with a two-week release cycle, content couldn't keep up. Every new feature shipped with placeholder microcopy. I needed to build a system, not just write better copy.
Research & Discovery
- Analyzed drop-off data - 68% of users abandoned at step 4 of 12
- Reviewed 6 months of support tickets to identify most-asked "how do I" questions
- Interviewed 8 users who completed onboarding and 6 who churned in the first week
- Audited competitor onboarding flows to identify patterns and gaps
Finding 1 - Value Too Late
Users needed to complete 4+ steps before experiencing core value. Churned users consistently said "I couldn't figure out what it was for."
Finding 2 - Feature-First Guidance
Every tooltip explained the interface element. None explained what the user was trying to accomplish or why this step mattered.
Finding 3 - One-Size-Fits-All Failure
The onboarding assumed a single user type. In reality, 4 distinct personas had completely different goals and entry points.
The Content System I Built
1. Goal-Based Onboarding Architecture
Redesigned the onboarding flow around user goals, not feature checklists. Instead of "Complete your profile ? Connect integrations ? Invite team," the new structure was: "What do you want to achieve first?" - three paths based on primary goal, each reaching core value in 3 steps or fewer.
2. Persona-Responsive Microcopy Framework
Built a content framework with four persona variants for key onboarding touchpoints. The system used 3 signals (role, team size, primary goal) to serve contextually appropriate guidance. Same interface, four different content experiences.
3. AI-Assisted Drafting Workflow
Built a prompt library and content templates that let me generate first drafts for new feature guidance in ~40% of the normal time. The system included: a voice and tone prompt, persona context blocks, and a QC checklist I ran every draft through before publishing. AI handled volume; I handled quality.
"She built a system that made our content team feel like three people. And the onboarding actually explains what the product does now, which is new."
- Product Manager Feedback4. Content Governance for a Fast-Moving Roadmap
Created a content review protocol integrated into the engineering sprint process. Every new feature shipped with: reviewed microcopy, in-app guidance, and a help article stub. No more placeholder copy going to production.
Results
This project demonstrates what AI-assisted content design actually looks like in practice - not AI replacing UX writers, but UX writers building systems that use AI responsibly. The prompt library, QC checklist, and governance protocol outlasted my engagement and continued producing quality content after I left.