# How does this actually work?

- An AI agent that reads the customer's existing website, content, and data and compiles it into structured Buyer Context Assets — the content layer that powers every discovery session. The wizard synthesizes offering information, messaging, personas, industry segments, use cases, buyer contexts, and messaging overlays into a reusable, structured library.
- A managed repository of all generated and refined buyer context assets. Customers generate, refine, and publish the context that powers Sandfox responses from this surface. Teams manage context assets and discovery maps, deploy experiences across channels, and read engagement signals by account and session.
- The buyer-facing layer. When a buyer arrives from any channel (website, paid, outbound, follow-up), Sandfox interprets their context and renders a full adaptive page — not a chat thread, not a static landing page — that reshapes as the buyer explores. Each buyer builds their own Discovery Graph (their unique path through the Discovery Map the customer's team defines). The experience adapts in real time to the buyer's role, company, history, and areas of interest.
- Every session is captured as structured buyer signal across four types: Buyer signal (what the buyer is trying to solve, questions asked, differentiators that move them, readiness), Account signal (which accounts are evaluating, who in the buying committee is participating), Experience signal (which entry points and paths drive progression, CTA performance, drop-off points), and Market signal (which messaging and content resonates, which personas engage most, content gaps). The Discovery Graph captures how decisions actually get made and accumulates into a living understanding of how buyers evaluate.
- The operator-facing control surface. Described as 'one admin surface' from which customers manage context assets, configure discovery maps, deploy experiences across channels, and read engagement signals by account and session. Dashboard metrics include: links minted, visitors, engagement rate, topic views per visitor, account match rate, sessions over time, top topics explored, top personas identified, top industry segments, and questions with no supporting asset.
- Buyer Context Assets are published as an MCP (Model Context Protocol), making the structured context a portable layer any agent in the customer's stack can use. Any AI agent in the customer's stack can draw from the same context layer that Sandfox compiles.