News
OpenAI dots: what these always-on AI agents change for an SMB
OpenAI introduced dots on September 29, 2026: always-on AI agents in ChatGPT. A dot helps one person; it does not replace an agent wired into your tools.
3 min read
The SynAI team
Key points
- OpenAI's dots, introduced on September 29, 2026, are always-on AI agents in ChatGPT, each with its own cloud computer.
- A first dot is included in the ChatGPT Pro and Business Premium plans; OpenAI has not published the price of additional ones.
- A dot suits one person's work; a process that touches your customers needs an agent wired into your tools, with human approval.
What did OpenAI announce?
OpenAI introduced dots at its DevDay on September 29, 2026. According to OpenAI, each dot:
- gets a goal rather than a question, moves forward on its own and brings its work back for review;
- runs on GPT-6 Astra, on its own cloud computer;
- connects to more than 4,000 apps and answers in ChatGPT, Slack and Teams;
- looks for ways to help in the background, read-only;
- follows your rules: act on its own, ask before acting, or hand off to you.
What does it change for an SMB?
A dot helps one person with the tasks they handle alone:
- preparing a meeting from the web and your documents;
- tracking a competitor or a regulation;
- writing the first draft of a proposal.
It has three limits.
- It works for one person. It acts with its user's apps and permissions, with no email address of its own and no announced API; dots for organizations are in pilot projects.
- Its memory cannot be managed item by item. According to OpenAI, you cannot view or erase what it remembers one item at a time: you have to delete the dot.
- Personal information. In Quebec, Law 25 requires a privacy impact assessment before personal information is communicated outside Quebec, and your customers have a right of access and correction. This is not legal advice.
A dot or an integrated agent: which one?
A dot serves one person; an integrated agent serves a company process.
| A dot | An integrated agent | |
|---|---|---|
| Who it serves | One person | A team, or your customers |
| Where results go | To ChatGPT, Slack or Teams | To your CRM or your phone system |
| Data | Public or not sensitive | Customers' personal information |
| Control | Your rules in ChatGPT | Human approval, a log of every action |
Who it serves
- A dot
- One person
- An integrated agent
- A team, or your customers
Where results go
- A dot
- To ChatGPT, Slack or Teams
- An integrated agent
- To your CRM or your phone system
Data
- A dot
- Public or not sensitive
- An integrated agent
- Customers' personal information
Control
- A dot
- Your rules in ChatGPT
- An integrated agent
- Human approval, a log of every action
Should you try a dot now?
Yes, if your plan includes one, on a personal task, for two weeks:
- Pick a repetitive task, such as preparing your meetings.
- Connect only the apps it needs, and forbid it to send emails.
- Do not give it any customer's personal information.
- Track the time saved and the mistakes, then keep it, extend it, or move to an integrated agent.
Frequently asked questions
- Are dots available in Canada?
- Yes. As of October 2, 2026, OpenAI is rolling them out on ChatGPT Pro everywhere except the European Economic Area, Switzerland and the UK, and on Business Premium wherever ChatGPT is available.
- Can a dot send emails on my behalf?
- Yes, if you connect your email account and your rules allow it. According to OpenAI, an automatic review then checks each email before it goes out.
