What rules does Dots ship with?

Built-in defaults on when to act alone and when to ask, custom rules per user to allow, block, or require approval for specific actions, an auto-review step for actions that affect accounts or share information, and a few tasks like password changes that always stay with the user. Background research runs read-only. The gap is that the rule lives with each user; the company needs one written once.

Why does an owner matter more than the safety settings?

Because the settings decide what a dot may do and the owner decides whether its output is true and useful. Thirteen of my 39 agents died with nothing referencing them, not because they misbehaved but because nobody was accountable for what they produced. A dot that learns preferences and picks up projects in the background is built for scope drift, which was the biggest predictor of death in that fleet.

What should a company measure after enabling Dots?

How many dots still complete production work after ninety days, and what each completed outcome costs. A dot per user is agent count with a cloud computer attached, and agent count is a vanity metric. Ask for the ninety-day number before approving the second dot.

Who Should Own an OpenAI Dot Inside a Company

Who should own an OpenAI dot inside a company?

THE SHORT ANSWER

A named human, per dot, before it is switched on, and the name goes on the dot's output. OpenAI Dots gives every Pro and Business Premium user an always-on agent with its own cloud computer that does read-only research in the background and acts when asked; specialist dots carry company credentials. Of the 39 agents I shipped in 80 days, 13 ended up orphaned and the best predictor of survival was a named owner. Underneath the owner sits one org-wide stop rule, written by product and not left to per-user settings: money, contracts, individual customers, and compounding decisions require a human approval the user cannot turn off. A dot with no owner is turned off, not adopted.

OpenAI shipped Dots on 2026-09-29, and the org question arrives with it: the agent department from The Triad Is Dead. Pods Are Dead. Agent Departments Now. is now a plan feature any employee can switch on. OpenAI Dots: The Agent Department Is Now a Line Item sets three rules for product by Friday: an owner per dot, from the orphan count in 39 PM AI Agents Deployed: What Stuck, What Died, and Why; one org-wide stop rule, from The Stop Rule; and a designer on the approval and recovery surfaces.

It belongs to the Product Leadership argument: the org chart was calibrated to a build cost that collapsed, and the work now is deciding which roles stop being necessary and which new ones, like the dot owner, appear.

Last reviewed 2026-09-30 · 1 min read

Who should own an OpenAI dot inside a company?

SOURCES

THE LONG VERSION

RELATED ANSWERS