Twelve things teams actually do once their AI reads the brain. Each one is a job someone used to do by asking around, digging through Drive, or rewriting the same document. Each comes with the prompt that starts it and what a good result looks like. Prompts are written for a chat client; in a coding agent, say the same thing.
1. Onboard a new hire in an afternoon
Who: the new person, on day one, with the brain connected.
I just joined as an account manager. Using the brain, give me: what the company does in three lines, who is who in my department and who I report to, the three most recent decisions that affect my work, and the pages I should read this week. Link every page.
A good result names people pages, the department hub, and decisions with paths. What used to take a week of meetings takes twenty minutes and is correct, because it comes from what the company wrote down.
2. Prepare for a client call
Who: anyone in sales, delivery or leadership.
I have a call with Vierci tomorrow at 10. Pull everything the brain has on them: the relationship owner, what we sold or proposed, open items, last contact, and anything sensitive I should know. End with three questions I should ask.
The AI walks the organization page, its links to deals and people, and the change history. If a page is restricted to another team, it says so instead of guessing.
3. Write the weekly team update from what changed
Who: a team lead, Friday afternoon.
Draft this week's update for the product department from the brain's changes in the last seven days. Group by project, name who made each change, and flag anything that reads like a decision. Keep it under 300 words.
The brain's change history is authoritative (who, what, when), so the update is built from facts rather than memory.
4. Find out why something was decided
Who: anyone who inherited a decision.
Why did we choose custom pricing for company accounts instead of a public price? Give me the decision page, the date, who decided, and what it superseded.
Decisions are first-class pages in the brain, with dates and supersession links. The answer cites the page, so the discussion can continue from the record instead of from recollection.
5. Find the owner of anything
Who: everyone, all the time.
Who owns the relationship with our payment provider, who is the backup, and which department does it sit in?
Systems, organizations and people are linked. The AI follows the links rather than searching text, so "owner" means the person the brain says, not the person mentioned most.
6. Produce a document in house style with a skill
Who: anyone who writes proposals, decks or reports.
List the skills in the brain. Then run the proposal skill for a Company Brain engagement with Acme: two departments, twelve people, kickoff in October.
A skill is a saved procedure the AI runs the same way every time. The team keeps the format, the pricing components and the tone in one place, and stops rewriting them per document.
7. Turn meeting notes into brain pages
Who: whoever ran the meeting.
Here are my notes from today's steering meeting. Extract the decisions, the open questions and the owners. File each decision as a decision page in the right department and add the open questions to the project page. Show me what you filed before you finish.
On a company brain, changes can be held for review depending on the department's settings, so a reviewer confirms before the rest of the company sees them.
8. Bring a Drive folder into the brain
Who: an admin or a department lead, with a Drive connector next to the brain connector.
Look through the "Client proposals 2026" folder in my Google Drive. For each document with lasting information (not drafts, not duplicates), summarize it into the brain with a clear title, a link to the source, and the client it belongs to. List what you filed.
The AI reads through its own Drive connector and writes through the brain. LAN never sees the raw files or your Drive credentials.
9. Write the board or investor update
Who: the CEO or chief of staff.
Draft the monthly investor update from the brain: revenue and pipeline from the finance and sales hubs, the two biggest product changes, one risk, one ask. Cite the page for every number. Flag any figure older than 45 days as needing a refresh.
Numbers that go stale are the main risk in updates. The brain marks volatile figures so the AI can flag them instead of repeating them.
10. Give your coding agent the company's technical context
Who: engineers using Claude Code, Cursor or VS Code.
Before you touch this service, read the brain's software section: conventions, the decision about API versioning, and the system card for our database. Then review my open file against them.
The software section of a brain holds codebase docs, conventions and technical decisions. An agent that reads them first stops reinventing patterns the team already settled.
11. Check a policy or process before acting
Who: support, operations, finance.
A client is asking for a refund three weeks after purchase. What is our policy, what exceptions exist, and who approves them? Give me the page.
Processes are pages too. The answer comes with the exact wording the company adopted, not a plausible paraphrase.
12. Teach it something durable
Who: anyone, the moment they catch a mistake.
That is wrong: Caro Carrasco moved from operations to partnerships in August. Update the brain and tell me what you changed.
With the instructions block in place, the AI folds durable corrections into the right page and reports the change in one line. The next person who asks gets the corrected answer, in every client.
Where to go next
- Good habits: how to ask so the brain does the work.
- Security and access: what the AI can see when it does all of this.