OpenAI has announced Dots, proactive AI agents designed to handle tasks with minimal supervision, at its Dev Day event in San Francisco. These always-on systems can work around the clock, preparing invoices or integrating data, and interact with over 4,000 apps via plugins. New Slack and Microsoft Teams integrations allow users to converse with their Dot outside of ChatGPT and Codex.

Dots operate using their own cloud computer and browser, with options to connect to your PC for direct control. At launch, Pro, Business, Premium, and Enterprise users in select markets get one primary Dot, customizable and nameable. Future plans include controlling teams of Dots.

Use cases include scientists integrating lab data into papers and developers collecting user feedback to write bug fixes. Dots can be messaged or called within ChatGPT on desktop, web, and mobile, carrying context across surfaces.

Safety features include using saved passwords without exposing them to the model, and read-only tools for proactive research, preventing unauthorized actions. Users can manage permissions and set custom approval rules. Dots roll out today to Pro and Business Premium subscribers in select markets, with the first Dot free and usage limits applying in Codex and Work.