The four parts of a Company Brain

Description and use

Implementing the four‑part Company Brain model enables AI to persistently store, prune, and expose knowledge, preventing redundant learning and ensuring accurate answers.

Problem

AI systems lack persistent, structured knowledge, causing them to rediscover information and produce inconsistent answers.

Outcome

Implementing the four-part Company Brain reduces AI’s need to relearn, ensures consistent answers, and keeps knowledge current while allowing agents to access relevant facts efficiently.

Workflow

• 1. Collect signals from sources (e.g., Sentry alerts, Slack threads, calls) and add them to an inbox folder.
• 2. Store each fact as a single, dated note in a structured folder (e.g., decision_, feedback_).
• 3. Run a dream/prune routine to merge duplicates, retire outdated facts, and archive old memories.
• 4. Provide agents with a door (e.g., a shared folder or API endpoint) so they can query the structured notes within their context window.

Evidence

AI is rediscovering knowledge from scratch on every question. There's no accumulation.

Source

The brain has to hear what happened: calls, chats, error reports, emails.

Source

A fact becomes memory when it sits in a shape you can ask about: one fact, one place, a date on it.

Source

Dreaming is the brain tidying itself: a job that re-reads what was learned that day and re-links it.

Source

Agents need a door. People ask questions in a chat. Agents need the facts delivered to them.

Source