OpenAI has introduced Dots, a new framework of always-on, autonomous AI agents designed to execute complex, multi-step workflows on behalf of users with minimal human supervision.
OpenAI has introduced “Dots,” a new class of always-on artificial intelligence agents designed to operate continuously, handle multi-step workflows and execute tasks independently on a user’s behalf around the clock.
Unveiled at DevDay 2026, the technology marks a major evolution from reactive chatbots into proactive, autonomous digital assistants. Powered by OpenAI’s flagship GPT-6 Astra model, each Dot functions on its own dedicated cloud computer complete with its own web browser. The system can connect to over 4,000 application plugins including productivity, coding, and communication tools allowing it to manage complex, long-term background projects without requiring constant human monitoring.
Unlike traditional AI assistants that only respond when prompted, Dots operate continuously in the background to learn user preferences, adapt to feedback and execute scheduled tasks.
When users are not actively interacting with them, Dots conduct what OpenAI terms “proactive research,” analyzing connected apps to identify helpful actions such as drafting forgotten invoices or organizing schedule timelines. Users can communicate with their assigned Dot via text or voice across desktop, web and mobile versions of ChatGPT as well as third-party workplace channels such as Slack and Microsoft Teams. Because context is retained across platforms, a project initiated inside ChatGPT can be picked up and updated within a Slack thread without re-explaining background details.
To address privacy and safety risks associated with autonomous software agents, OpenAI built explicit security guardrails and approval frameworks into the platform’s architecture.
A Dot’s cloud computer is kept isolated from a user’s local machine by default, though users can grant explicit permission for local file access. When operating autonomously in the background, Dots are limited to read-only access across connected applications, preventing them from sending outbound communications or altering files without user verification. Critical or consequential actions—such as modifying passwords, making financial transactions, or transferring sensitive records require explicit human approval through customizable control rules.
OpenAI is making the first Dot available to Pro and Business Premium subscribers at no additional charge, while rolling out enterprise pilots for corporate teams to automate specialist functions across procurement, customer support and commercial operations.
1 thought on “OpenAI Unveils “Dots”: Always-On AI Agents That Work on Your Behalf”