Designing for Generative AI

Modern AI user experience (UX) design is shaped by the AI’s ability to learn and evolve, enabling dynamic interfaces that respond in real time to each user’s unique context. For instance, AI-powered algorithms analyze behavior and situational data to anticipate needs—adjusting layouts, content, and interactions even before a user makes a request. Effective Generative AI design must balance automation with user control.

As multimodal interfaces evolve, hands-free interactions through voice and gesture are becoming more intuitive. AI systems must seamlessly switch between input modes based on context and user behavior, even adapting tone by recognizing emotional cues to foster more human-like interactions. Designers are now optimizing AI-driven interfaces not just for usability and performance, but also for accessibility, emotional intelligence, and contextual relevance.

This demands a rethinking of traditional UX principles, as user interactions become more dynamic, intelligent, and adaptive. AI integration in design is shifting the focus from static layouts to fluid, realtime experiences that adapt across devices and contexts. The result is a new era of hyper-responsive, evolving digital ecosystems—interfaces that continuously learn from and grow with every user interaction.

Our Design Approach

At our studio, we believe designing for AI isn’t just about enabling intelligence—it’s about making that intelligence accessible, responsible, and relatable. We take a thoughtful, human-centered approach that ensures AI experiences are intuitive, transparent, and aligned with user needs—not alienating or opaque.

After all, what good is a smart system if it’s as confusing as quantum physics on a Monday morning?

Informed by deep user research and insights from our senior design team, here are five key principles we apply when designing AI-driven products:

AI “hallucinations” occur when systems generate responses that sound plausible but are factually incorrect or misleading. As designers, we tackle this challenge by crafting interactions that promote transparency, user awareness, and critical thinking. We build in friction and feedback loops that help users question, verify, and improve responses.

Design example:

A health assistant might clearly state, “This is not a substitute for professional medical advice” and include source links or confidence indicators for each recommendation—building trust while setting boundaries.

2. Aligning with Mental Models

We invest in understanding users’ mental models—what they know, expect, and how they make decisions. This allows us to design AI systems that align with familiar behaviors, enabling smoother onboarding and more meaningful interactions. Equally, we help AI systems form useful models of the user.

Design example:

By interviewing frequent online shoppers, we learn how they browse, compare, and filter products—then design a conversational shopping assistant that mirrors these patterns and preferences.

3. Supporting Rationalization and Context

We design interfaces that empower users to interpret and rationalize AI outputs rather than accept them blindly. This includes surfacing relevant data points, showing reasoning, and enabling “why” behind “what.”

Design example:

A financial assistant doesn’t just say “invest in X,” but explains why—citing data trends, market sentiment, and associated risks—so the user feels informed, not instructed.

4. Accounting for Model Limitations

We embed AI limitations into the experience intentionally—not as disclaimers, but as essential design elements. This helps set appropriate expectations and fosters long-term trust.

Design example:

A financial forecasting tool reminds users that predictions are probabilistic, not guarantees, and may miss recent market shifts. This prevents overconfidence and invites user discretion.

5. Integrating Domain Expertise

Collaboration with domain experts is core to our process. Whether we’re designing for legal, medical, or financial applications, we ensure the AI is grounded in field-specific knowledge and upholds industry standards.

Design example:

In a legal assistant, partnering with compliance experts helps surface jurisdictional nuances and allows the system to flag when a professional review is needed—avoiding misinterpretation.