Defining a visual and interaction language for the AI-first expert platform

AI Native IEP is the foundation of a new generation of expert experiences — a GenAI-powered platform that changes how experts handle calls and cases. It marks a paradigm shift away from static, navigation-heavy systems toward a fluid, anticipatory interface where information, tools, and insights arrive at the right moment. The challenge: codify that vision into a design language that feels alive, intelligent, and distinctly AI-native — not as a theme or aesthetic, but as a philosophy of interaction. How might we express the intelligence, adaptability, and humanity of AI Native through its design language? The target qualities: Fluid — adaptive, responsive, always in motion. Intentional — purposeful hierarchy through behavior and anticipation. Human — trustworthy, empathetic, emotionally resonant. Intelligent — guided by context and foresight, not just data. Rather than commit early, we explored several distinct creative directions to find what AI Native could look, feel, and behave like.

Salt is essential yet invisible: it enhances everything around it, brings balance, and reveals depth without taking center stage. Like a chef balancing flavors, the modern expert understands that great work comes from harmony — and AI is the understated enhancer that elevates human ability without overpowering it. The visual language leans on balance and restraint, with material cues of glass, translucence, crystalline structure, and shimmer.

Neo Mind makes the language tangible as "fluid intelligence." The expert moves between customer chat and a parallel conversation with Intuit Intelligence, and the prototype maps every state in between — customer chat, AI chat, and the two guidance states where AI surfaces real-time suggestions without interrupting the flow. The switch between talking to a customer and talking to AI is made deliberately obvious so a message is never sent by accident.

Lucid pursues the same state model with a quieter, more restrained surface — "human intelligence, enhanced." Guidance and the AI conversation recede until the expert reaches for them, and the single "Ask anything" field keeps the interface calm. It tests how far the language can hold back and still feel intelligent and present.

Arc puts the conversation and the work side by side. The chat thread sits next to a live document canvas, so research, drafting, and the customer exchange share one workspace. This direction tests the "information always at hand, never in the way" principle at full density — keeping context visible without crowding the expert.

A second Arc pass pushes the canvas further — customer chat with canvas and AI chat with canvas — exploring how the expert pivots between conversation and a working surface as the task shifts. It shows the layout flexing between modes while the underlying language stays consistent.

The expert copilot prototype brings the language into a full working shell — engagements and customer rails on the left, a structured workflow, and an "Ask AI" entry point woven into a real TurboTax expert conversation. It's the system in context: identity, interaction patterns, and components composed into the surface an expert would actually work in.