Company second brain
What Doram knows about your company, where it comes from, and how it stays current.
The company second brain is what your company knows about its data: what things are, how they connect, and how each number is counted. Every AI system reads from it through one path, starting with Doram's AI analyst. It's stored as a semantic context graph.
Where it comes from
Doram reads four kinds of source, always with read-only access.
| Source | What Doram learns from it |
|---|---|
| Databases | Schema, column profiles, query history, and definitions already in dbt and BI tools |
| Business context | Terms, decisions, goals and calendar rules, entered by owners because no system holds them |
| Company knowledge | Rules, terms and decisions from documents in tools like Notion and Confluence |
| Third-party tools | Events and reference lists from tools like Jira, Slack, your CRM and billing |
How it's built
Every fact goes through three steps:
- Discover. The crawler agent drafts the fact from evidence. A draft without evidence goes no further.
- Validate. Code checks the draft. Metrics and joins are run on real data. Decisions and policies are checked for evidence and conflicts.
- Approve. Low-risk changes go live on their own. Anything that could change an approved number waits for its owner.
Trust levels explains each step in detail.
How it stays current
After the first build, the second brain maintains itself. A monitoring agent watches every source for changes and redrafts anything a change breaks. A feedback agent learns from how people use the analyst, and acts only when at least three people, or one owner, point the same way. Every change goes through the same three steps, and users keep the last approved version until the new one is approved.
What it doesn't store
No rows of your data. The second brain holds definitions, relationships, column profiles with a few masked sample values, and the evidence behind each fact. Your data stays in your warehouse.