AI-driven campaigns that target the right audience at the right moment
Founding designer on this AI marketing platform. Built the onboarding, campaign creation, and content approval flows from an early admin console into a shippable v1. The platform scored 72.1 on the System Usability Scale (above the industry average of 68).
2026
Year
AI/ Marketing Automation
Category
Figma + MCP | ClaudeCode | FigmaMake | GitHub
Stack
Problem
Solution
Trade-offs & decisions
Approval friction vs. AI training speed. Content generated by the AI required marketer approval before sending, carried over from the admin console. This added friction to the marketer's workflow, sometimes felt like extra manual work on top of campaign setup. We kept it anyway because with only one pilot client, approval and rejection data was the main signal available to train the AI's judgment.
It was a decision that made sense for a single-client pilot with limited data, not necessarily the right call at scale. With more clients and more data, I'd expect this to loosen into something like auto-approve with an audit trail, but we never got the client base to test that assumption.
Testing & Iteration
Two moderated usability rounds early in the project, one with 4 participants and one with 6, run directly with marketers at the pilot client (a banking company). Sessions were one-on-one, screen-share based, using specific mockups to walk through proposed flows and changes.
Alongside moderated sessions, I used Inspectlet to track how users actually navigated the admin console: where they hesitated, where they dropped off, which flows caused friction. This wasn't a formal quantitative study, but it surfaced real usage patterns that fed directly into flow changes, separate from what users told us in interviews.
A third round happened post-launch: a moderated in-person session in Amsterdam with 7 participants, marketers and managers from the pilot client, run by a design consultant who joined the team in its final three months. This session was recorded, produced a findings document, and included a follow-up survey that generated a System Usability Scale score.
Next Steps
That's the end of this project.
Wanna hop to the next?
prototype blocks tested