Redesign how your company works
Mirror your current org in Kylon: departments, people, AI employees, and who sees what.
Mirror your current org in Kylon: departments, people, AI employees, and who sees what.
Your business is mapped. Now build the frame: where departments go, how people and AI employees are staffed, and who sees what.
An AI-native company gives AI employees a clear place in the org.
No need to reorganize for Kylon.
Tip: the owner of the business is Owner, the tech lead is Admin.
Each Room also has its own admins and members. More: kylon.io/help/member-and-admin-permissions
Codex can also run on Kylon Cloud with a ChatGPT subscription, computer off. More: kylon.io/help/bring-your-own-agent
The platform enforces all four. No prompt gets an Agent around them. Agent memory comes last.
Public is open, private is invite only.
Each Agent has its own access list.
Each App sets its own access.
Whoever connects an account owns it.
More: kylon.io/blog/how-permissions-work-in-kylon
Example: the Sales Agent sits in the Sales Room. Ops is public, so it joins when asked to look. Finance is private, so it can't read it.
Example: the Finance Agent is for the finance lead and the owner only. The sales lead shares a Room with it and still can't use it.
Every Agent needs at least one human admin. Workspace admins not on the list can't use it either. More: kylon.io/blog/agent-governance-hard-boundaries
Example: the customer list is for sales and ops. Payroll is a separate App for the finance lead and the owner.
Apps with a database can't be public. An App can only use connections granted to it.
Example: the sales lead keeps their inbox to themselves, the booking calendar is company-wide, and the team Slack goes to ops and the Ops Agent.
"Everyone in workspace" means people only. Add Agents one by one. Long-term Agent access to someone's connection needs the owner's sign-off.
Example: the Finance Room notes a customer's annual fee was waived. A salesperson asking the same Agent usually won't get that.
More: kylon.io/blog/agent-governance-hard-boundaries
Ask first: if this ended up in the Agent's own memory, would it be a problem?
Mainly Google Cloud, in the US.
Pro and Ultra plans never train on your content.
The four permission layers decide.
More: kylon.io/help/where-data-is-stored kylon.io/help/is-my-content-used-for-training kylon.io/help/what-agents-can-access
Workspaces shared with colleagues stick far more often.
The owner, one or two colleagues, one Room.
Once it runs smoothly, invite the whole department.
Other departments follow, all in one workspace.
Then fill in the table on the next page.
Which are private? Which need full isolation?
Who uses each Agent? Who manages it?
Which tables are sensitive? Who gets each?
Whose account? Shared with whom?
| Room | Public or private | People in it | AI employee, who uses it | App, who opens it | Tools, who connects |
|---|---|---|---|---|---|
| e.g. Sales | e.g. Public | e.g. Sales lead, two reps | e.g. Sales Agent, whole company | e.g. Customer list, sales and ops | e.g. Everyone connects own inbox |
| e.g. Finance | e.g. Private | e.g. Two finance staff, owner | e.g. Finance Agent, these three only | e.g. Payroll, these three only | e.g. Accounting software, finance lead connects |
Draw from your current org chart. No need to reorganize for Kylon.
More: kylon.io/help/switch-workspaces kylon.io/help/invite-or-remove-members kylon.io/help/what-agents-can-access
Where departments go, how people and AI employees are staffed, who can see what.
Your first colleagues are already in.
Like hiring anyone new: give AI employees a role, access and approvals.