Build Agent Adoption Signals
The first ask pointed toward quick UI fixes to boost Build Agent. Before recommending any, I wanted to know what the existing data already said. I pulled together CSAT feedback, Build Agent telemetry, and past research, then ran a heuristic evaluation across two real workflows: a new user looking for a course, and an experienced developer troubleshooting their instance.
- Secondary research
- CSAT feedback
- Build Agent telemetry
- Heuristic evaluation
Developers weren't struggling with Build Agent. They were struggling to orient themselves on the portal at all. Adoption was the symptom, and orientation was the cause.
🧭 The rationale
Why start with existing data
Interviews take time and budget. Analyzing prompt logs, CSAT surveys, and past case studies first surfaced the biggest friction points: fragmented navigation and weak search, outdated and buried documentation, and no clear starting point.
Why combine sources
Each source answered a different question. Data showed where developers dropped off, and the design review explained why. Together, they turned friction themes into concrete recommendations.
Working with thin signals
Of the 2,094 survey responses I analyzed, 600+ came in after Build Agent launched, and only one of those named it directly. Rather than overstate what indirect signals could prove, I treated Study 1 as directional: it showed us where to look, not the full picture.