Lennar's first AI data agent — enterprise data made conversational
BigLen is Lennar's internal LLM-powered data assistant. As Founding AI Designer, I shaped the product's design foundation from early UX exploration to production-ready experiences — a complete interaction system for AI-powered data work: conversational analytics, guided analyses, scheduled automations, and a visible Trust layer that makes every AI-generated insight verifiable down to its sources.
Four rules shaped every decision: plain English first, guide before generate, trust must be visible, and scalable by design. From there the work ran Discover to Launch — strategy grounded in data-driven insights, definition of the UX AI workflow, high-fidelity UI and prototyping in tight iteration loops, a design system refined through feedback, and product handoff to engineering through GitHub.
Enterprise data lived behind dashboards and analysts. The goal was to make it conversational for non-technical teams without sacrificing trust — an LLM answer is only useful for business decisions when you can see why to believe it: which agents ran, which enterprise sources ground it, and how fresh the underlying data is.
Shipped a complete interaction system for AI-powered data work: plan mode, conversational analytics that turn a question into structured business output (charts and KPI summary cards), guided analyses, prompt templates, scheduled automations with email delivery, and conversation branching with context snapshots. Every answer carries a Trust Score — data freshness, quality, grounding, and execution health — with source-level clarity. The design system ships with interactive documentation for consistent engineering adoption.
End-to-end design as the founding designer: conversational AI workflows and interaction patterns, high-fidelity UI across mobile and desktop in light and dark themes, UX prototyping, the Trust layer experience that makes AI reasoning inspectable, and the full design system — tokens, components, and interactive documentation — handed off to production through GitHub.