OpenAI launches dots, always-on AI agents for work

OpenAI has introduced a new AI product called dots, rolling it out to Pro, Business Premium and Enterprise users in eligible markets.

Dots are always-on software agents designed to handle tasks on behalf of users through ChatGPT, Slack and Microsoft Teams. Each dot runs on its own cloud computer, can connect to more than 4,000 apps through plugins, and can work across multiple projects at the same time.

The launch reflects OpenAI's broader push beyond chatbot-style interaction toward persistent assistants that keep working in the background. Users can give a dot a name, connect apps, set permissions, and communicate with it by message or voice.

The product is built on GPT-6 Astra. OpenAI says dots can learn from feedback over time, retain context across channels, and carry out work using their own browser and connected applications.

How It Works

Users can inspect a dot's work at any time by opening its cloud computer. A dot can also be allowed to connect to a user's laptop, where it can work directly alongside them.

OpenAI positions the product as a way to delegate ongoing tasks rather than manage separate chat threads. A dot can continue working on several assignments while new tasks and ideas are added.

Examples outlined by OpenAI include software development, product launches, scientific research, sales proposals and content production. In one example, a dot monitors customer feedback, prepares software fixes, tests them, and produces pull requests for review. In another, it updates launch materials when product scope changes. Other examples include rerunning research analysis as new data arrives, revising sales proposals as requirements change, and drafting creator content from interview transcripts.

Enterprise Focus

Alongside personal dots, OpenAI also disclosed a preview of specialist dots for organisations. These are intended to take on defined responsibilities such as access management, IT-provisioned hardware, and deep connections to company systems of record.

Specialist dots would operate with their own identities, credentials and system access, while organisations would define their responsibilities, tool access and approval processes. Early internal testing has covered procurement, invoice processing, email marketing, customer support and commercial contracting.

OpenAI is starting with focused enterprise pilots. It is also working with Microsoft to integrate specialist dots with governance and security controls in Microsoft Agent 365, aiming to let businesses manage the agents through existing Microsoft tools.

Controls And Safeguards

Dots include additional privacy and safety measures on top of existing ChatGPT protections. Users can choose which apps a dot can access, manage permissions through ChatGPT controls, and set custom rules to allow, require approval for, or block actions.

For background “proactive research” when users are not actively working with them, dots use read-only access. Under that arrangement, the agents can review connected app data but cannot send messages, change content, or control a browser or computer.

For more sensitive actions, dots use an automatic review system to check whether a task fits user instructions, custom rules and safety requirements. Certain tasks, including password changes, remain restricted to the user. Monitoring systems can also pause or stop a dot's work if they detect a safety concern.

On data use, OpenAI does not use content from ChatGPT Business, Enterprise or Edu workspaces to improve its models by default. It also does not train directly on proactive research activity or a dot's internal notes, although information from those processes may be used in some eligible conversations or tasks depending on account settings.

Pricing And Access

A first dot is included at no extra cost with Pro and Business Premium plans. Enterprise users, including Edu and Healthcare customers, can access the beta if a workspace administrator enables it.

Conversations with a dot do not count toward standard ChatGPT usage limits. However, tasks initiated through Codex or ChatGPT Work do count toward those limits.

Users can create their first dot through the ChatGPT desktop app or a desktop browser, then continue conversations through the mobile app after setup. “Dots can still make mistakes, so always review consequential work,” OpenAI said.