Designer Fund: how Sierra's design team supports a 20x engineering organization
Sierra is a startup building AI agents for enterprise customer service. It reportedly passed $200 million in annual recurring revenue within two years. Its design team is small enough that it cannot personally review every product decision in an engineering organization roughly 20 times its size. This case study from Designer Fund, the first in its AI in Design 2026 series, describes how that imbalance shaped the way Sierra’s designers work.
The core shift is from screen-fixing to system-building. Early on, designers tried catching and correcting engineering decisions one at a time — refining a table layout here, adjusting a component there. The team concluded this approach didn’t scale and had a quieter cost: designers who spend most of their time on quick corrections don’t build the judgment that comes from debating design decisions in depth. Instead, the team moved to building shared infrastructure. One example is consolidating roughly 18 variants of data tables into a single flexible component that any team can extend. That work multiplies across the product.
Prototyping with real data replaced wireframes and static mockups. Designers sketch in Figma, then build working versions in Cursor connected to actual production data. Designer Wayne Fan describes the rhythm as “cook, prototype, be wrong, go back.” The goal is to surface problems that static screens hide before they reach engineering.
The team also invests time in certain hand-crafted, low-efficiency work. For the Ghostwriter agent’s icon, designers spent hours refining a 16×16 ghost figure and animating it in Cursor. The mascot became a team rallying point. Fan frames this as a conscious decision: not all slow work is wasted time, especially when it builds quality standards and team identity.
For designers working inside large organizations, or on small teams supporting large engineering groups, the Sierra case offers a specific reframing: the question isn’t which screens to fix faster, but which shared foundations will reach the most of the product. AI tools expand how much a small team can prototype and iterate, but the more durable gain comes from deciding what to build at scale.