Connie was built to serve one of the more demanding environments in conversation design: a voice assistant designed to help employees navigate their 401k - retirement savings decisions that are, for most people, genuinely high stakes.

Designed for a major US enterprise, Connie supports employees in managing their retirement accounts through natural conversation. The subject matter demanded a careful balance: the assistant needed to feel approachable and easy to talk to, while handling financial information with precision and care. Getting that tone right - helpful without being casual, clear without being clinical - was central to the design challenge.

Under the hood, Connie runs on a vector database, enabling it to retrieve and surface relevant information fluidly across a broad and complex knowledge base. Designing for this architecture meant thinking carefully about how responses were structured and how the assistant handled the edges of its own knowledge - especially important in a financial context where vague or incorrect answers carry real consequences.

Connie represents some of the most demanding work in this portfolio - designing for a context where clarity isn't just good UX, it's a matter of financial accuracy and employee trust. It was a project that required not just strong conversation design instincts, but a genuine understanding of the stakes involved for the people on the other end of the conversation.