Switchboard

Switchboard is a data platform. It ingests and stores raw data from your sources. You model business logic on top of that data. Then you run reporting, chat, and automations against the result.

Underneath it is a data warehouse built on open storage formats. Around the warehouse sit features that make it easy to build and run data workflows. The point is to run a whole data stack in one tool instead of several, and to be AI native from the ground up.

The four stages

Connect. A source is a connection to a system you already use. Shopify, HubSpot, an ad platform, a spreadsheet. Switchboard syncs it on a schedule. Each source delivers resources, which are the tables that connection provides.

Store. Everything lands in your warehouse. It is included with Switchboard and managed for you.

Model. Resources arrive in the shape the source vendor chose. Models reshape them into your business's terms. What a customer is. What counts as revenue. Which order was somebody's first.

Use. Four ways to work with modeled data:

• Documents are pages that combine text with live charts and tables.
• Operator is the built-in AI. You query, analyze, and build by asking.
• Automations run work on a schedule or a condition.
• The MCP connection exposes the same models to Claude.

Every one of those reads the same models. A metric therefore has one definition across the platform. Operator reads those definitions rather than inferring your business rules from column names.

Setup and what comes after

Sources and a starting set of models and documents are put in place during implementation.

After that you extend them. Ask Operator for a new metric. Add a block to a document. Schedule a report.