OpenAI released always-on AI agents called Dots

OpenAI launched Dots, a new class of always-on AI agents, during the DevDay 2026 event. These agents act as digital assistants that stay active in the background to complete complex tasks for users.

Powered by the GPT-6 Astra model, Dots can crawl the web and pull information from connected software. Instead of waiting for a single prompt, they monitor context and proactively suggest actions. Imagine a scenario where your agent spots a scheduling conflict and immediately provides food delivery options from GrubHub based on your calendar.

Users can interact with these agents through ChatGPT, Slack, or Microsoft Teams. For those on the Pro subscription, which costs $100 per month, there is also support for iMessage and RCS on Android.

Enterprise focus

OpenAI also introduced specialist versions of these agents tailored for business roles. These tools are designed to handle specific professional workflows like marketing, accounting, and legal analysis.

To keep things secure, Dots require explicit user permission for sensitive tasks like changing account passwords or installing new software. Users can also set custom rules to restrict what their agents can or cannot access.

While this rollout marks a shift toward proactive assistance, the system still has boundaries. The agents are not yet able to handle multiple simultaneous complex projects for every user, and their effectiveness depends on the quality of the data users connect to them. Are you comfortable giving an AI agent permission to manage your daily logistics?