OpenAI’s Dots explained: How its ‘always on’ AI agents work, what they can do

OpenAI has unveiled its line of always-on personal AI agents called 'Dots', unveiled during its annual developers' conference on September 29. CEO Sam Altman explained that Dots can handle everything, from booking restaurant reservations to complex multi-step processes like invoicing. Powered by GPT-6 Astra, each Dots agent operates within its own virtual machine (VM) or cloud computer, relying on browser access and plugins to connect to over 4,000 apps.

Dots learns from user feedback and works towards assigned goals 24/7, always striving to achieve tasks. While it has faced setbacks during live demos, Dots is designed to be proactive, conducting research on connected apps when not assigned a task, but it is restricted to read-only mode to prevent any unauthorized actions.

Unlike many other personal AI agents in the market, Dots presents its work in a way that is tailored to the user, offering suggestions before they even think of asking. The agent allows users to personalize it by naming it and customizing it to their preferences. Users can communicate with Dots through chat or call, even on Slack or Microsoft Teams apps shortly. Once a project or task is assigned, Dots begins working immediately and can add new tasks without managing separate conversation threads.

Dots differs from other agents by its ability to carry over context across various channels, meaning users can start a project in ChatGPT and continue it on Slack. The AI agent is designed to help OpenAI secure its lead in the race to build the next-generation of personal AI agents. It is currently available for ChatGPT Pro and Business Premium subscribers in eligible markets, with enterprise users having access to the beta version.