OpenAI's dots: what always-on AI agents mean for a small business
OpenAI's dots are AI agents that keep working between chats. What they are, who can use them, what they cost, and how a small team should start.
On September 29, 2026, at its DevDay event, OpenAI introduced "dots": AI agents that keep working after you close the chat window. You give a dot a goal, connect the apps it needs, and it carries on in the background, checking in when a decision needs you. For founders and small teams, it is the clearest sign yet that AI is moving from a tool you visit to a coworker that is always on.
Here is what dots are, who can use them today, what they cost, and how to try one without creating a mess.
What a dot actually is
OpenAI describes dots as always-on agents. A few details set them apart from a normal ChatGPT conversation:
- They run on their own computer. Each dot has its own cloud computer and its own browser, so work continues even when your laptop is off. You can open that computer at any time to see what it is doing, and take over if you want.
- They connect to your apps. Through OpenAI's plugin ecosystem, OpenAI says dots can connect to more than 4,000 apps. A dot can only use the apps you have connected and the permissions those connections allow.
- They remember and learn. A dot keeps its own notes on your preferences, decisions and ongoing work, and OpenAI says it learns from your feedback over time.
- They are reachable where you work. You can message or call your dot in ChatGPT on desktop, web and mobile, and in Slack and Microsoft Teams. OpenAI says texting is coming soon.
- They juggle several projects. One dot can run a project while working on others, without you managing separate threads.
OpenAI also shared a small, very relatable example: an early tester's dot noticed he had forgotten to invoice a client, prepared the invoice, and sent it after he approved it.
That last part, "after he approved it," matters most.
How approvals and permissions work
Dots are designed to ask before they act on anything that matters. According to OpenAI's announcement and its documentation:
- Background work is read-only. When you are not actively working with it, a dot does what OpenAI calls "proactive research" using read-only tools. Those tools cannot send messages, change content in your apps, or control a browser or computer.
- Risky actions get reviewed. Before an action that could affect your accounts or share information, an automatic review checks it against your instructions, your rules and OpenAI's safety requirements. The result is that the action goes ahead, waits for your approval, or is handed back to you.
- You can set your own rules. Custom Rules let you allow specific actions, require approval for them, or block them outright. Built-in safety requirements always apply on top.
- Some things always stay with you. OpenAI says certain sensitive tasks, such as changing a password, are never done by the dot.
- Passwords stay hidden from the model. For supported websites, a dot can use saved passwords without the model seeing them, and using a saved login for a new sign-in needs your confirmation.
- There is an activity log. An Activity View shows what your dot is doing, including background work.
OpenAI is also candid that dots can make mistakes, and its documentation notes that stopping a task does not undo actions already completed. Review anything consequential before you approve it.
Who can use dots, and what they cost
Dots are rolling out gradually, and access depends on your plan:
- ChatGPT Pro: available to users over 18, except in the European Economic Area, the UK and Switzerland.
- ChatGPT Business Premium: rolling out worldwide in supported regions.
- Enterprise, Edu and Healthcare: a beta that is off by default until a workspace admin turns it on.
Free, Go and Plus plans were not included at launch, though OpenAI says it plans to expand to more users.
On cost, OpenAI says your first dot is included in Pro or Business Premium at no extra charge. Plans include an allowance for deeper work, with extended limits for the first month after launch. Conversations with your dot do not count toward ChatGPT usage limits, but tasks it starts in Codex or ChatGPT Work do.
OpenAI has said you will later be able to add more dots and pay to increase a dot's speed or monthly workload. It has not published prices for those add-ons yet, and plan prices vary by region, so check OpenAI's pricing page for your country before you budget.
Setup happens on desktop; after that, you can message your dot from the mobile app too.
Good first jobs for a dot
The best first tasks are repetitive, low risk and easy to check. Some examples that suit a small business:
- Inbox and follow-up triage. Summarize what came in overnight and draft replies for you to approve.
- Invoice and payment reminders. Flag unpaid invoices and prepare reminders, the same pattern as OpenAI's own example.
- Meeting prep. Pull together notes on a client before a call from your CRM, email and shared files.
- Competitor and market watching. Track changes on a short list of websites and send a weekly digest.
- Content drafts. Turn a call transcript or product update into a first draft of a post or newsletter.
Notice the pattern: the dot does the gathering and drafting, and a person makes the call.
Guardrails to set on day one
Treat a dot like a new hire with access to your systems. Before you connect anything, decide:
- Which apps, and at what level. Connect only what the first task needs. Read access to your CRM is a very different decision from permission to send email as you.
- What always needs approval. Anything that sends money, emails customers, deletes data or changes account settings should require your sign-off.
- Whose data it can see. On personal plans, you can control whether dots' conversations and work are used to improve OpenAI's models. OpenAI says it does not use content from Business, Enterprise or Edu workspaces for training by default. If you handle client data, a business plan with admin controls is the safer starting point.
- Who reviews the log. Pick one person to check the Activity View weekly, at least at first.
Where a custom integration helps
Plugins cover popular apps well. Many small businesses also run something less standard: a custom CRM, an internal database, a booking system, or a line-of-business tool built years ago.
For those, an agent is only as useful as the doorway you give it. A small, well-scoped integration, such as an API that lets an agent read open orders but not delete them, is often safer and more reliable than handing an agent a login and hoping it navigates your admin screens correctly. OpenAI is also piloting "specialist dots" with large organizations: agents with their own identity, credentials and access to core systems. That is a hint at where this is heading, and clean, permissioned access to your data will matter more each year.
What this means for you
You do not need to rush. If you are on an eligible plan, try one dot on one narrow job for a few weeks, with approvals switched on for anything that leaves your company. Measure whether it saves real time. Meanwhile, look at your own systems: if an agent could not safely reach your data today, that is worth fixing whether or not you use dots.
If you want help connecting AI agents to your CRM, website or internal tools with the right permissions, talk to SEVA Systems. We build custom software and integrations designed to be safe for both people and agents.
Sources
- https://openai.com/index/introducing-dots/
- https://learn.chatgpt.com/docs/dots
- https://chatgpt.com/pricing