Introduction
I led the design of a new subscriber-focused web app for Envato, built to bring our stock and AI tools into a single, cohesive experience. The project focused on unifying fragmented workflows, removing friction from core interactions, and setting the platform up to introduce new subscriber pricing tiers.
The impact was immediate across our North Star metrics:
- Downloads per user increased 20%.
- Downloads of AI generations rose from 30% to 35%.
- Subscriber CSAT climbed from 4.2 to 4.5.
The challenge
Previously, Envato ran a single web app for guests, free users, and subscribers alike. There was no dedicated subscriber experience.
Meanwhile, our AI tools had been built rapidly and ad hoc by different teams on a completely separate stack. This created significant architectural constraints and a fragmented user experience. We needed to design a subscriber app that could seamlessly integrate stock and AI, while giving us the flexibility to roll out new features and pricing tiers more easily.
The approach
Make the core loop fast
We trimmed everything unnecessary from the core loop (search and download stock) and focused on making it fast to use. We introduced features that were previously identified as being high impact, such as infinite scroll, better content recommendations from the same author, one-click downloads, and other quality of life improvements that speed up user workflows.
Stock & AI integration
We integrated AI tools into the same app architecture and design language, creating consistency across every tool in the workspace. Generated content now lives alongside stock assets in a single workspace, so subscribers can save, organise, and access everything in one place.
Ship quick, iterate faster
The project also reoriented the team around agentic coding, which changed how we design and build at a fundamental level. Design became more code-first, production cycles shrank dramatically, and we increased continuous interviewing practices to validate decisions faster. The result was a team that could ship, learn, and iterate in much tighter loops.
The result
