Use AI Prototypes to Test Complex Designs Earlier

Complex interfaces have many states and possible user actions. Filters, dashboards, and conversational AI are difficult to show with a few static screens. In the past, teams often tested only simple paths and waited for developers to build the product. AI tools now help designers create working prototypes earlier. These prototypes can include realistic data, complex filtering, and changing responses.

At Ramp, Pavan Garidipuri used Cursor to redesign an expense-policy editor. The experience allowed administrators to edit rules through AI chat. Static screens could not show the changing conversation well. Test participants opened a working prototype with sanitized versions of their own data. They interacted naturally and found that the way edits were tracked and displayed was confusing. The team might have missed this problem without a realistic prototype.

Another Ramp designer, Andrew Lucas, built a conversational expense-reporting prototype. The connected agent could answer many different inputs and produce changing results. Participants were free to type unexpected requests instead of following one planned path. This helped researchers see how the agent behaved in less predictable situations. Such problems often appear late when teams test only fixed screens.

Purdue University researchers also compared static wireframes with an AI-generated traffic-data interface. Their dataset had 27 attributes, including nested details. Static screens helped users discuss visual design, but they could not truly use the filters. Later, the team built an interactive version with v0, Bolt, and synthetic data. Participants gave more specific feedback and found practical problems. The study was not a controlled experiment, so it does not prove AI always creates better research.

AI tools execute design decisions quickly, but they do not replace design skill. Without clear context and guidelines, they may choose poor patterns or weak visual hierarchy. A polished prototype may also look ready for production when it still contains hidden assumptions and errors. A skilled designer should review it before development. The main benefit is earlier learning: teams can test realistic behavior before spending the full cost of building the product.