Skip to main content
AI Tools8 min read

ChatGPT Agents: What Changes When Dots Switch On at Work on 18 November

R

Raina Rufus, Founder of xSquads

Published

ChatGPT agents, in November 2026, are three OpenAI products under one name: the ChatGPT agent (agent mode), workspace agents a team can share, and dots, always-on agents with their own cloud computer. On 18 November 2026, dots in eligible ChatGPT Enterprise, Edu, and Healthcare workspaces become available to members by default unless an admin opts out, and billing starts for enabled dots. One personal dot per member is a generalist. The hard part is which agent owns which task.

What ChatGPT agents are now

ChatGPT agent, or agent mode. The overview on OpenAI's ChatGPT agent help page says: "ChatGPT agent is no longer available." It points you to ChatGPT Work for longer, multi-step tasks. Further down, the same page still describes agent mode: select it from the tools menu or type /agent, on Pro, Plus, Business, Enterprise, and Edu. The file comparison is ChatGPT agent alternative. ChatGPT Work is ChatGPT Work alternative. Workspace agents. The workspace agents page says they are in research preview for ChatGPT Business, Enterprise, Edu, and Teachers plans. You describe the job, connect apps and files, and share the agent so a team follows the same workflow. The page says agents can run on a schedule, and that admins set role-based access, audit logs, and approval gates for steps such as sending a message. Dots. A dot is an always-on agent in ChatGPT. The getting started guide says you give it a goal, it can keep making progress between conversations, and it has its own cloud computer and the apps you connect. You choose what it can do on its own, and you can pause it. The 4 November 2026 release notes say personal dots can continue working on tasks between conversations. The file-based comparison is OpenAI Dots alternative. A plugin is a different object, a package of skills and apps: ChatGPT plugins.

What changes on 18 November 2026

On 4 November 2026 OpenAI made dots generally available to eligible ChatGPT Enterprise, Edu, and Healthcare workspaces, with an admin preview until 18 November 2026. The release notes say owners and admins can enable access or opt out before dots become available by default on 18 November. Dots are not available for workspaces with UAE data or inference residency. The same notes say new dots support data residency and Enterprise Key Management, with inference residency in the EU and the US, and that dots are HIPAA-eligible in covered workspaces.

Through 18 November 2026From 18 November 2026
Who can use a dotA member whose role an admin has enabledMembers, by default, unless the workspace opts out
Does access create a dot?No. Each member chooses one personal dotSame
Price of the dotOpenAI says the preview is freeBilling begins for enabled dots
Work or Codex threadsBilled separately at standard ratesSame

The getting started page and the admin guide are the sources for that table. Access does not create a dot: each member creates one personal dot. Enabled dots are dots that are not paused. Enterprise can pay flat rates by type of work, or scaled consumption. A Work or Codex thread is a separate bill, and Enterprise model controls do not apply to the main dot.

@theaicatchup flagged the default and the billing start for Wednesday 18 November. @agentsnacks_ai flagged the admin preview through 18 November and the opt-out after that date.

Custom instructions, and when a dot must ask

OpenAI's what's new page adds custom instructions for a dot in the 9-13 November 2026 update, set in chat or under Customize, then Behavior. The controls page says they are separate from ChatGPT custom instructions and cover how the dot works, how often it checks in, and when it asks before acting. OpenAI's example is: "When you finish a task, link to the result and tell me what needs my attention."

@alexmichael posted: "show me a draft before sending any emails." Gadgets360, updated 13 November 2026, reported the same pattern: prepare an email and show the draft before it is sent.

Custom rules are for a specific action. The controls page lists four:

RuleWhat the controls page says
Take action without askingTake the specified action without asking for approval.
Take action when you say soProceed when you explicitly request the action. Otherwise ask immediately before acting.
Ask before taking actionAsk for approval before taking the specified action.
Hand off to youAsk you to take the action instead.

The page's examples are Ask before taking action for customer messages, and Hand off to you for deleting shared files. A rule does not grant app or computer access, a draft request does not authorise a send, and a dot can still make mistakes. If Use custom rules for dots is off, saved rules do not apply.

Several computers, Europe reports, and Wallet

@nxtday wrote on 10 November, citing @OpenAIDevs on 9 November, that one dot can connect to several local computers and that you pick which machine handles each task. The getting started page says local access starts turned off: connect a computer in the desktop app and confirm Allow access. Changing the selected computer does not move an existing task. Allow local computer access is a separate admin opt-in, and enabling dots does not connect a computer.

On 13 November, @AGTPinsights wrote that users across the EU reported seeing dots in ChatGPT that day. The getting started page still says Pro dots roll out to users aged 18 and older in markets excluding the European Economic Area, Switzerland, and the UK. Those are user reports, not an official UK launch.

@JoshExile82 asked OpenAI on 9 November to let a dot reply to bot and app wake-up mentions in Slack, and wrote that it does not work, so their apps and bots cannot wake the dot. On the same day as the Europe reports, OpenAI's ChatGPT release notes introduced ChatGPT Wallet in dots for ChatGPT Pro subscribers aged 18 or older in the United States.

One dot per person is not a team plan

From 18 November, every member of an eligible workspace can switch one dot on unless an admin opts out. Activity for that person sits under In progress, Scheduled, and Completed. Admin Console, Dots, shows who created a dot, whether it is active or paused, and whether Slack or Teams is connected. That list does not answer "Show me what they finished."

Ten agents doing nothing is just ten tabs. Multiple AI agents help when each task has an owner and a place the team can read the result. Single agent vs multi-agent is when to split a job. AI teammates is roles and handoffs. AI agent workflow is the path a task takes. Microsoft's control plane is a different product: Microsoft Agent 365.

What to set before and after 18 November

  1. Decide who gets Use dots. Under Admin Console, Permissions, Workspace capabilities, turn the workspace default on only if every member should have it. Otherwise leave it off and grant a custom role. Check every role. The admin guide says if one role grants access, turning it off in another role does not remove it.
  2. Set Add dots to Slack and Microsoft Teams, Allow local computer access, and Use custom rules for dots on purpose. Local access is a separate opt-in. The permission does not install Slack or connect a computer.
  3. Use Ask before taking action for customer sends, and Hand off to you for deleting shared files. Test a harmless action. A written policy is not a product restriction.
  4. Add the finish line to custom instructions: when a task is done, link to the result and say what needs attention. Add the draft-before-send line if email is in scope.
  5. Give each task one owner, and keep one place for Done and In review. OpenAI's profile is per person. A shared board is a separate decision.

xSquads for Work

xSquads for Work gives every knowledge worker their own squad of specialised agents working together as a swarm, an alternative to chatting with ChatGPT or Copilot. Each task has a clear owner on one board, from Backlog to In progress to Review to Done. The desktop app is live on Mac and Windows and is free with Pro agents. The product is at xsquads.ai.

xSquads Software Factory

A dot can hand engineering work to a Codex thread, and OpenAI bills that thread separately. For an engineering team, xSquads Software Factory is the line for code. Engineers run coding agents in the cloud that raise pull requests from user stories. Review agents check each pull request against the intent of the story, not just passing CI, and request fixes. Self-improving loops feed lessons back into the next run. How that line runs is in the software factory post. The product is at xsquads.ai.

Frequently asked questions

What are ChatGPT agents?

OpenAI uses the name for more than one product. The overview on the ChatGPT agent help page says ChatGPT agent is no longer available and points to ChatGPT Work for longer, multi-step tasks. The same page still describes agent mode, started from the tools menu or by typing /agent. Workspace agents are a research preview on Business, Enterprise, Edu, and Teachers plans: build one agent and share it across the workspace. A dot is an always-on agent in ChatGPT with its own cloud computer, and it can keep working between conversations.

What changes for ChatGPT Enterprise on 18 November 2026?

On 4 November 2026 OpenAI made dots generally available to eligible ChatGPT Enterprise, Edu, and Healthcare workspaces, with an admin preview until 18 November 2026. From 18 November, members have access by default unless the workspace opts out, and billing begins for enabled dots. Each member can create one personal dot, and turning access on does not create it for them. Work a dot delegates to ChatGPT Work or Codex is billed separately at standard rates. Dots are not available for workspaces with UAE data or inference residency.

Can I turn off dots in ChatGPT Enterprise?

Yes. In Admin Console, under Permissions and Workspace capabilities, Use dots is the switch. Keep the workspace default off and grant access with a custom role if only some people should have it. Check every role on a member: if one role grants Use dots, turning it off on another role does not remove access. An admin can also pause a personal dot from Admin Console. An admin pause stops the member resuming it, and billing stops while the pause lasts.

How many dots can each person have?

One. In eligible ChatGPT Enterprise, Edu, and Healthcare workspaces, each member can create one personal dot. The member chooses whether to create it. Access alone does not create a dot.

Is one ChatGPT agent enough for a team, or do you need several?

One personal dot is a generalist for one person. A team still needs an owner for each task, a place where finished work shows up, and a person who reviews it. Multiple AI agents do not answer that on their own. xSquads for Work gives each knowledge worker a squad of specialised agents on one board, from Backlog to In progress to Review to Done, with one owner per task.

Where these facts come from

Read on 17 November 2026.

Related reading

Raina Rufus, Founder of xSquads

On 18 November a member can switch on one dot unless an admin opts out. A squad is how each task comes back with an owner. Get xSquads.

Give yourself a squad of AI teammates

Your own squad of specialised agents to delegate to and review. The app is live on Mac and Windows. xSquads for Work is free with Pro agents.