The client
A B2B SaaS company headquartered in Bangkok, around forty employees, selling a project-management tool to professional services firms across Asia. Their trial-to-paid conversion had plateaued at around fourteen percent, which their head of growth was confident was below the ceiling for the product.
The problem
The trial onboarding was a hardcoded seven-email sequence, the same for every signup. It introduced features in a fixed order regardless of what the user was actually doing inside the product. Power users got told basics they had already figured out. Stuck users got sent feature announcements when they actually needed help with the thing they were trying to do right now.
The approach
The build wired Segment behavioral events into n8n. Every meaningful action a new user took (or notably failed to take) inside the product became a signal. Instead of a fixed sequence, the system held a model of where each user was stuck and what they had already mastered. An LLM drafted the next email based on that state: a contextual nudge, a specific tutorial link, an offer of a fifteen-minute call with someone on the team.
The product team kept full editorial control over content libraries and tone. The dynamic part was which content went out, in what order, and exactly when.
The result
Trial-to-paid conversion climbed from fourteen percent to twenty-eight percent over the first three months. Time to first meaningful action inside the product dropped, which was a signal the head of growth had been chasing for a year. The customer-success team also reported that the calls they did get were better-prepared, because the right users were now reaching out with specific questions instead of confused requests for a general demo.
"Honest, scoped, on-time. The thing was running before the deadline they had committed to. I have worked with three other AI vendors before this and none of them shipped anything we kept using past month two."