Remember when "AI assistant" meant a chat box that forgot you once the tab closed? OpenAI just gave its assistants their own cloud computers, a browser, and a to-do list that keeps going while you sleep. The intern has stopped waiting to be asked.
Meet the Agents That Don't Clock Out
At DevDay 2026 in San Francisco on September 29, OpenAI launched Dots, always-on AI agents inside ChatGPT. They run on GPT-6 Astra. Each Dot gets its own isolated cloud computer and virtual browser, connects to more than 4,000 apps, and works in ChatGPT, Slack and Microsoft Teams. You give it a goal and it keeps working toward that goal across several projects without needing a nudge at every step.
OpenAI's launch example: one tester's Dot noticed an invoice had never been sent to a publication, prepared it, and sent it once approved. Actions that touch accounts or share information go through an "auto-review" check, and OpenAI warns that agents "can still make mistakes." The rollout includes one Dot per plan for ChatGPT Pro and Business Premium subscribers, with Enterprise, Edu and Healthcare in beta. Pro subscribers in the EEA, Switzerland and the UK are left out for now.
Delegation Is a Governance Problem Now
The real change isn't the model, it's who's in charge of the task. A chatbot answers a question, but a Dot owns an outcome, and that turns "what can it access?" from an IT footnote into a board-level question. OpenAI clearly knows this: enterprise specialist agents get their own identities and credentials, and Microsoft handles governance through Agent 365.
Most businesses will skip the boring part, which is that an agent is only as useful as the systems it can safely reach. If your CRM, CMS and invoicing tools are held together with spreadsheets and goodwill, a 24/7 agent will just make mistakes faster. The companies that win with agents will be the ones that cleaned up their data plumbing first.
Dots are a real preview of how work gets done next, so make sure yours has a job description before it starts sending invoices.
Thinking about which of your workflows an always-on agent could actually take over, and which systems it would need permission to touch? Let's sketch that out together before you hand anything the keys.
Source: SiliconANGLE