The ReDuX design team resisted the temptation to build what was technically possible. Instead, we started by watching architects work—shadowing them during legacy assessments, observing how they sketch on whiteboards, and listening to their frustrations with existing tools.

One pattern emerged repeatedly: architects don’t think in flat diagrams. They think in layers. They zoom out to see ecosystems. They drill down to see components. They filter by type. They rearrange spatially to match their mental models. They ask questions aloud—”What does this database connect to?” or “Show me just the external dependencies.” That’s when it clicked: their natural workflow alternates between visual and verbal. They look at the diagram (canvas), ask a question out loud (conversation), someone points at a box (canvas), someone else explains the relationship (conversation). The whiteboard captures the visual, but the understanding happens in the dialogue.

So we designed for both. Not as separate tools, but as complementary modes that reinforce each other.

These observations became design principles:

Progressive disclosure over information overload. The hierarchical C4 structure—System Context, Container, Component, Code—mirrors how architects naturally move between altitude levels. You start high and dive deep only when needed.

Spatial thinking matters. Architects understand systems through arrangement. The drag-and-drop canvas isn’t decoration; it’s fundamental. When you move a component closer to a database, you’re not just organizing pixels—you’re building understanding through spatial relationships.

Conversation reveals intention. Sometimes the fastest interface isn’t visual at all—it’s language. But sometimes language alone is too abstract. The breakthrough was realizing these aren’t competing approaches—they’re complementary. The agentic AI chat doesn’t replace the visual canvas; it works with it. You see a complex diagram, you ask: “Show me all databases in the payment system.” The canvas highlights them visually. You drag them into a group. The AI understands the spatial arrangement and can now answer questions about that cluster. Visual thinking + conversational inquiry = understanding that neither could create alone.

Different lenses, same system. The C4 lens is one of several—ScreenFlow, Code Insights, API Insights—because architects need multiple perspectives on the same truth. No single view captures everything. Design’s job is making movement between lenses seamless and intuitive.

Prototyping with Purpose

Our team built and rebuilt the C4 lens through rapid cycles of Figma Make prototypes and user testing. But we weren’t testing features—we were testing understanding.

Early sessions revealed surprising failures. Architects got lost when drilling into containers because they lost context of which system they were inside. The solution: always show parent boundaries. This small design choice—drawing a subtle border indicating the parent system—became fundamental to preventing disorientation.

Another discovery: color matters more than anticipated. Assigning consistent colors to entity types (pink for Person, blue for System, teal for Container, purple for Component, green for Database, gray for External) turned scanning into pattern recognition. Architects could spot all databases across a complex diagram instantly, without reading a single label.

The AI conversation design went through the most iterations. Early versions felt like chatbots—separate, toy-like, disconnected from the architectural work. Users said explicitly: “I don’t want another chat interface. I want something that feels like a knowledgeable assistant.”

The breakthrough came from treating the AI as a design partner, not a search box. The conversational interface became contextual—appearing as a drawer tied to the canvas, with grouped action buttons that signaled intent (analyze, explore, suggest, reimagine). The conversation design provided architecture insights, suggestions, bulk actions, deeper analysis of entities, prompt library, and contextual actions.

Testing revealed that the small details—button states, hover feedback, transition timing—built trust. If micro-interactions felt polished, users trusted the AI’s architectural suggestions. If they felt janky, users questioned the underlying intelligence, even when the suggestions were sound.

This is the truth about designing for technical users: they’re not more forgiving of poor design. They’re less forgiving, because they know how good design should work.

Designing the Conversational UI

The AI conversation interface went through the most iterations. Early versions felt like chatbots. One architect said: “I don’t want to chat with my architecture. I want to interrogate it.”

We made it contextual—a drawer attached to the canvas with action buttons (analyze, explore, suggest, reimagine). Conversation and canvas aren’t separate features—they’re two modes of the same exploration. You see something, you ask about it. The AI highlights it visually. You drag components together, the AI understands the grouping. Visual and conversational, working together.

Lessons from Designing for Architects

Building the C4 lens taught the team principles that apply beyond ReDuX:

1. Design for disappearance. The best interface is the one users forget they’re using. Every design decision should ask: does this help the work, or does it get in the way?
2. AI needs a voice, not just an interface. Conversational AI can’t feel like a feature tacked on. But it also can’t exist in isolation from the visual canvas. The magic happens when they inform each other: you see architecture visually, you interrogate it conversationally, the visual updates based on your questions, you ask new questions based on what you see. It’s a feedback loop, not two separate features. This means visual integration, contextual awareness, and design polish that signals they’re one unified tool.
3. Progressive disclosure isn’t dumbing down. Hiding complexity until needed isn’t condescending—it’s respectful of human attention. Show what matters now, reveal depth on demand.
4. Spatial interaction reveals mental models. Letting users drag, arrange, and organize isn’t just convenient—it externalizes their thinking. Watch how someone reorganizes a diagram, and you see how they understand the system.
5. Trust comes from craft. Technical users notice typography, spacing, color consistency, animation timing. These details aren’t cosmetic—they signal that you care about quality at every level.

This is part one of our three part series on using ReDuX C4 visualizations for digital transformation. Read part one here: When Architects Stop Fighting Their Tools.

To see ReDuX in action, book a demo with our team!