Designing with AI agents inside a real UX workflow

A practical AI-assisted UX process at Economatica: competitor research, website audit, Obsidian documentation, Figma synthesis, and recommendations grounded in human product judgment.

At Economatica, I used Claude agents as research and critique partners to investigate competing financial platforms, review the new website experience, and turn findings into concrete interface and content recommendations. The value was not that AI generated design decisions. The value was that AI expanded the research surface while I remained responsible for prioritisation, product context, and final UX judgment.

The challenge

Economatica operates in a dense financial software category, where credibility, precision, and trust matter. The new website had to explain products, integrations, APIs, data workflows, and use cases without becoming generic fintech marketing. The risk was a site that looked polished but still carried avoidable UX and content issues: inconsistent terminology, repeated imagery, unclear affordances, missing visual support, and page behaviours that could weaken confidence.

What changed

This work created a more systematic way to move from broad AI-assisted exploration to actionable UX critique. It helped separate category conventions from opportunities for clearer Economatica messaging, and it made the designer's judgment more visible: what mattered to financial users, which patterns were worth borrowing, and which recommendations actually fit the product.