Data layer
Where your data lives, how it is classified, and why no agent ever holds it all.
The Data layer is where your company's knowledge, records and documents live, and where their classification is applied. It is the layer most platforms ask you to compromise on. Adam AI does not.
What it holds
- Company knowledge. Documents, policies, procedures, correspondence and the structured records agents reason over, indexed on your box.
- Operational data. Tickets, contacts, invoices, jobs: the records your systems of record hold, reached through the Integration layer and cached only for the life of a session.
- Markings on data. Every object carries the Markings applied to it and the Markings it inherited. The Data layer enforces them on every read, not at the edge.
- Lineage. When an agent derives something new from marked sources, the derivation is recorded and the new object inherits every source Marking.
Where it lives
On hardware you own, or in the cloud account you own. Not in ours. Not in a model provider's. The box is yours; if you leave, the data is already in your building.
What it never does
- Data never leaves the estate to train a model. Models receive context for a session and retain nothing.
- No agent sees everything. A query returns only what the session's Markings permit, and derived results are filtered the same way.
- Nothing is silently redacted or deleted to make it "safe". Some platforms read your most sensitive disclosures in order to strip them out. We do not read them at all unless the holder of the Marking is asking.
- The dashboard's own database and the companies database are never mixed.
Why it matters
Agents are only as useful as the data they can reason over, and only as safe as the controls on that data. Putting the controls in the Data layer, rather than in each tool, means a new agent, a new skill or a new model inherits the same protection without anyone remembering to add it.
Related: People layer for Markings, Integration layer for systems of record.