Guides
OpenAI dots vs Harriet Agents. What Enterprise should know
Harriet Team · · 5 min read
On September 29, 2026, OpenAI launched dots, always-on agents that run on their own cloud computer and keep working between conversations. For a company, the first questions are which plan includes them and who controls them. Dots come with the $100 Business Premium seat, they are off by default on Enterprise, and they sit outside the workspace model controls.
What are OpenAI dots?
Dots are always-on agents inside ChatGPT. The details below are from OpenAI’s announcement of September 29, 2026 and its help center as checked on September 30, 2026.
- A dot is powered by GPT-6 Astra, OpenAI’s top model. It has its own cloud computer and browser.
- It connects to the apps a person has already connected in ChatGPT. OpenAI says its plugins cover more than 4,000 apps.
- It does background research without being asked. OpenAI says those research tools are read-only, so they cannot send messages or change content.
- Specialist dots, which get their own identity and credentials inside a company, are a preview. OpenAI is starting with enterprise pilots.
Which ChatGPT plans include dots?
OpenAI’s help center lists three routes on September 30, 2026.
Pro users get dots, except in the European Economic Area, Switzerland and the UK.
Business Premium users get dots in all supported ChatGPT regions. The standard Business seat is not on the list.
Enterprise, Edu and Healthcare workspaces get a beta. It is off by default until a workspace admin enables it.
How much do OpenAI dots cost?
OpenAI’s announcement says the first dot is included in a Pro or Business Premium plan at no extra cost, with an allowance for deeper work and extended limits for the first month.
So on Business, the price of a dot is the price of the seat. OpenAI’s pricing page on September 30, 2026 lists a premium seat at $100 a month on annual billing and a standard seat at $20. That is $80 a month more for each person you move up.
OpenAI says people will later be able to add more dots and scale their output, and gives no price. The Enterprise rate card did not mention dots on September 30, 2026, so how dot usage is billed on Enterprise is a question for your OpenAI contact.
Conversations with a dot do not count toward ChatGPT usage limits. Tasks a dot starts in Codex or ChatGPT Work count as usual.
What can a ChatGPT Enterprise admin control?
OpenAI’s admin guide covers Enterprise workspaces. On September 30, 2026 it describes these controls.
- A Use dots permission, set in the workspace default or in custom roles.
- Separate permissions for adding dots to Slack and Teams, for local computer access and for custom rules. Local computer access and custom rules are off by default.
- Cloud controls for browser use, network access and computer use.
Access is additive across roles. A role that grants dots still gives access even if another role turns it off, so removing access means checking every role a person holds.
Model controls do not reach dots. The guide says Enterprise model controls and default model settings do not apply to them. A workspace that steers people to a cheaper model has no such lever here.
We did not find an equivalent admin page for Business on September 30, 2026.
What happens to company data?
OpenAI’s announcement says content from Business, Enterprise and Edu workspaces is not used to train its models by default.
OpenAI’s privacy FAQ says a dot keeps context from conversations and connected apps for as long as the dot exists. Individual memories cannot be viewed or deleted. Disconnecting an app stops new access and does not remove what the dot already learned. Deleting the dot is the way to clear it.
The admin guide adds that a dot’s cloud computer does not inherit a member’s VPN, browser sign-ins or device policies.
OpenAI’s FAQ says its safeguards reduce the risk of prompt injection and do not eliminate it.
On Harriet, every skill and connector is reviewed, contained and audited before it can touch company data.
How do OpenAI dots relate to Harriet Agents?
They are the same kind of product. A Harriet Agent runs in a cloud workspace, and like a dot it keeps working when your laptop is closed. The differences are in who controls it.
| OpenAI dots | Harriet Agents | |
|---|---|---|
| Model | GPT-6 Astra only. Workspace model controls do not apply | Any model on the router. Admins limit which models each group can use |
| Identity | Specialist dots, in preview, get their own identity and credentials | Every action belongs to a named user |
| Memory | Context you cannot view or delete item by item | Files you can open and edit |
| Background work | Researches on its own, read-only | Runs the scheduled tasks you define |
| Apps | The apps connected in ChatGPT | The same reviewed skills and connectors as Harriet Desktop |
What should IT do this week?
- List who holds a Business Premium seat. Those people can create a dot as the rollout reaches them.
- On Enterprise, decide before you enable. Pick who gets dots, whether local computer access is allowed and whether dots can join Slack or Teams. Then review every role.
- Review connected apps. A dot works within the permissions people have already granted in ChatGPT.
- Add dots to offboarding. When someone leaves or changes role, deleting the dot is what clears its context.
Any agent that acts on its own raises the three questions a dot does. Who may use it, what can it touch, and what record does it leave?
Harriet answers them with access controls for each team, reviewed skills and an audit trail, across every model.
For how we approach a rollout, start with implementation. For the controls behind it, see tool safety and security. For what OpenAI’s top plan costs, read ChatGPT Enterprise pricing.
