Core concepts

Think of Semogram in two layers, like Palantir-style platforms do with data versus objects:

• Your data layer: the systems you already use — databases, warehouses, CRMs, files. Semogram connects to them but never replaces them. This is where facts live.
• The intelligence layer: what Semogram builds from those facts — structured entities, workflows, predictions — every piece traceable back to the data layer. This is where decisions get support.

Nothing in the intelligence layer is mysterious: each piece links back to the facts it came from. That traceability is the whole product.

The concepts

• Workspaces — what owns everything you do.
• Data and integrations — how your systems connect, and how syncing works.
• Workflows and plans — what a workflow is made of and how it stays valid.
• Ontology — durable entities and relationships in three layers.
• Runs and evidence — executions and the proof behind every output.
• Predictions and outcomes — kinds of forecasts and how results close the loop.
• Review and feedback — approvals, corrections, and how the system learns.
• Skills and guidance — reusable instruction that steers work.
• Semogram for LLMs — the compact version for AI assistants.

Image: two-layer diagram — data layer systems below, intelligence layer above, evidence links between.