Episode #93: How Generative UI Is Changing UX with Nick Babich

Nick Babich joins Dwayne Samuels to examine a different model for human-computer interaction. Instead of making people move among apps, icons, and windows, the team at brain.ai explored an interface that interprets a person's intent and presents the functionality needed in that moment. Nick calls it a morphing interface: AI powers the interaction, while the visible experience changes in response to a typed prompt or voice command.

The premise creates a demanding design problem. A single environment may need to support food ordering, travel planning, translation, ride sharing, or scheduling without becoming incoherent. Nick explains how his team identified important scenarios, found reusable patterns across them, tested unfamiliar interactions early, and measured whether people could complete tasks effectively. The conversation also covers design systems, collaboration, ethics, and the skills designers can develop as AI becomes part of everyday product work.

What you'll learn

• Why generative experiences should begin with user intent rather than a collection of screens

• How scenario mapping reveals which components can be reused across very different tasks

• Why design principles must come before styles, components, atoms, and molecules

• Which user and business metrics can expose problems in an unfamiliar AI interface

• How designers can incorporate AI tools into real workflows without surrendering human judgment

Design the system around intentions and scenarios

For Nick, the core challenge is not generating an interface. It is understanding the user's mental model well enough to provide an interface that solves the immediate problem. That shifts the design process away from treating each capability as a separate app. The team first defined the tasks its core users wanted to accomplish, then prioritized them against both user needs and commercial goals. Only after that did it identify the components required to support those tasks.

Some scenarios share interaction patterns even when their subject matter differs. Food ordering and travel planning, for example, can both use cards to present options and filters to narrow a list. Scheduling an event requires a different set of components. Mapping those scenarios and their intersections helped the team reuse elements instead of creating a new component set for every prompt. That restraint matters because an unconstrained component library can grow too quickly to remain manageable.

The same analysis informed work across touch and desktop interfaces. Not every mobile component transferred cleanly to the web, so the team adjusted components and looked for a useful intersection between platforms. Consistency did not mean making every experience identical. It meant preserving a coherent foundation while allowing each task and medium to demand an appropriate interaction.

Build principles first, then test the unfamiliar

Nick argues that a flexible design system starts with product strategy and design principles, not a component inventory. The team first clarified what it wanted the product to accomplish and why. It then defined foundational principles before building styles and components through an atomic design approach. Without those principles, a library may contain polished elements but still fail to scale or adapt when the product and market change.

Early research focused on behavior before visual polish. The team simulated the interface with a human operating behind the experience, allowing participants to interact as if the application were working. This exposed how people naturally expressed requests before the team committed to an implementation. The findings were analyzed and turned into interaction patterns and components, followed by repeated prototyping in tools such as Figma and Sketch.

Testing remained important because people outside the product team could respond very differently from those who designed it. The team tracked time on task and error rate for user performance. For first-time users, session length helped show whether they explored the available capabilities and found value. For returning users, task completion speed mattered more. Conversion rate and bounce rate supplied a business view. Nick also observed that participants were surprised that one interface could address many needs, and that they noticed crafted animation and transitions rather than valuing function alone.

Keep collaboration and judgment human

Morphing interfaces require designers, engineers, data scientists, sales, and marketing to share a clear view of what they are building, why it matters, and when it should be delivered. Nick warns against designing in a silo and treating engineering involvement as a final handoff. Instead, handoff activities should happen throughout the design process so technical constraints, user value, and commercial questions shape the work together. New teams may need especially frequent communication while they build shared context.

Nick expects design roles to evolve as AI speeds up parts of the process, but he does not see the need for human-centered design disappearing. People still decide what to build, release, and prioritize. His practical advice is to use AI tools consistently inside real design work, then learn which tasks benefit from them and which remain better served by established methods. That might include experimenting with Figma plugins for microcopy, layouts, or accessibility analysis rather than switching workflows entirely.

His own weakness turned strength reinforces the point. Nick can spend so much time learning from people that visual production starts later. Yet that research helps him design faster once he begins because he understands what must be built. In an AI-assisted process, clear problem definition remains more valuable than generating screens quickly.

Listen to Episode #93.