The problem at the core of Davis is a familiar one: working with large datasets is painful. Analysts jump between spreadsheet tools, query builders, and dashboards — none of them talk to each other, and none of them speak plain English.
We designed Davis as an AI-native chat experience that feels as familiar as ChatGPT but is purpose-built for data work. Users describe what they want in natural language — 'show me Q3 revenue by region as a spreadsheet' — and Davis handles the rest.
The interface was designed around three core interaction states: thinking (while the model processes), confirmation (scoping the request before committing), and delivery (presenting the result with a live preview). Each state has a distinct visual treatment so users always know where they are in the flow.
Spreadsheet generation was treated as a first-class feature. Users can preview outputs before confirming, drill into individual cells, and export directly from the chat — no context-switching required. The result is a data tool that feels like a conversation.