OpenAI used its annual DevDay conference on September 29 to unveil Dots, an always-on agent that can browse the web and operate external applications on a user's behalf, marking the company's clearest push yet toward autonomous AI assistants.
According to reporting from the event, Dots was presented as an agentic avatar — described by TechCrunch as "bubbly" — that works continuously in the background rather than only when prompted. The agent can use a browser and connect to outside apps, handling multi-step tasks that would previously have required repeated back-and-forth with the user.
Paid ChatGPT users began receiving access to Dots on the day of the announcement. The agent can be reached through ChatGPT itself and through Slack, with OpenAI indicating that additional channels are planned. The company also demonstrated a related feature, ChatGPT Space, during the event.
The launch places OpenAI squarely in the race to define what industry watchers call agentic AI: systems that act with persistence and autonomy rather than simply answering questions. Rivals including Meta and others have shown their own agent products this year, and the Barron's coverage of DevDay framed the announcements as OpenAI's bid to keep its lead in the consumer agent market.
For users, the practical pitch is that Dots can be given a goal — booking, research, monitoring, coordination — and work on it over time, checking back only when it needs a decision. For businesses, the Slack integration suggests OpenAI wants Dots embedded in everyday workplace workflows.
As with any always-on system, questions around permissions, data access and oversight will be central to how Dots is received. OpenAI has not yet published a full technical breakdown of the agent's guardrails, and reviewers will be watching how the company handles the security implications of an agent that can act in external apps.
Sources:
TechCrunch | Barron's
According to reporting from the event, Dots was presented as an agentic avatar — described by TechCrunch as "bubbly" — that works continuously in the background rather than only when prompted. The agent can use a browser and connect to outside apps, handling multi-step tasks that would previously have required repeated back-and-forth with the user.
Paid ChatGPT users began receiving access to Dots on the day of the announcement. The agent can be reached through ChatGPT itself and through Slack, with OpenAI indicating that additional channels are planned. The company also demonstrated a related feature, ChatGPT Space, during the event.
The launch places OpenAI squarely in the race to define what industry watchers call agentic AI: systems that act with persistence and autonomy rather than simply answering questions. Rivals including Meta and others have shown their own agent products this year, and the Barron's coverage of DevDay framed the announcements as OpenAI's bid to keep its lead in the consumer agent market.
For users, the practical pitch is that Dots can be given a goal — booking, research, monitoring, coordination — and work on it over time, checking back only when it needs a decision. For businesses, the Slack integration suggests OpenAI wants Dots embedded in everyday workplace workflows.
As with any always-on system, questions around permissions, data access and oversight will be central to how Dots is received. OpenAI has not yet published a full technical breakdown of the agent's guardrails, and reviewers will be watching how the company handles the security implications of an agent that can act in external apps.
Sources:
TechCrunch | Barron's
