Sierra’s Mossy gives voice agents a visible, living interface

A different approach to voice interfaces

Sierra created Mossy for its Sierra Summit event, where the company showcased voice capabilities with Tandem Voice and Horizon agents. The company wanted a visual element that could help people follow conversations, since much of what happens in a voice interaction is invisible.

Instead of using a glowing orb or pulsing waveform, Sierra designed a plant-like form. The choice reflects Sierra’s visual identity, which includes nature-inspired elements and a logo based on a pinecone seen from above.

The designer behind Mossy tested around 50 materials, including velvet, fleece, moss and grass. Moss came close to the desired look, but its dense surface made movement difficult to see. Grass offered individual strands that could bend and sway in response to sound. The name Mossy remained even after the material changed.

Code helped turn design ideas into a working system

The project was completed in two weeks alongside the designer’s other work. According to Sierra’s blog, AI coding agents were used to explore ideas, build prototypes and deliver the final design.

The designer built a tool with real-time controls for grass length, density, color, contrast, highlights, rotation, inertia and sound responsiveness. The tool also controlled how far movement overshot before settling. This allowed design changes to be tested immediately instead of being described to another person.

The process included pushing settings to extremes, from bald to fully overgrown, to identify when Mossy felt most alive. Sierra says AI made each iteration faster, while the design decisions still depended on human judgment.

Visual cues distinguish speaking from listening

Early feedback ranged from “It’s okay” and “Obsessed” to “Why does it look like a virus?” and “I can’t tell when I’m talking and when the agent is talking.” Sierra responded by softening the grass and adding small flowers, making Mossy look more like a plant.

The interface also uses different movements for each side of the conversation. When Mossy speaks, it rotates slightly and its grass moves in different directions. When Mossy listens, it pulls back and gently pulses, creating visual space for the person speaking.

Mossy was built in code from the beginning rather than being handed off from a Figma file or After Effects specifications. Sierra says that meant the design reached production without a separate translation step. The project shows how AI-assisted coding can let one designer explore and ship a branded interface more quickly, while leaving decisions about what works to the designer.