OpenAI introduced dots as AI agents for ongoing work

What OpenAI says dots can do

OpenAI describes the product as a way to give an agent an ongoing project rather than repeatedly request separate answers. Its examples include investigating software feedback, revising project materials and updating analyses as information changes. These are company-described capabilities, rather than an independent guarantee of performance. [1]

The announcement says the plugin ecosystem can connect dots to more than 4,000 apps. Access is determined by the accounts and permissions a user supplies; it does not mean every dot automatically has access to every application. [1] [3]

Who gets access, and what is included

OpenAI’s help page, checked on October 13, says availability is gradual. Pro access excludes the European Economic Area, Switzerland and the United Kingdom. Business Premium is available across supported ChatGPT regions, while Enterprise beta access requires a workspace administrator to enable it. [2]

The first dot is included without an additional charge in eligible Pro or Business Premium subscriptions. That is an inclusion in a paid plan, rather than a standalone free subscription. The plan also includes an allowance for more substantial work, with temporarily extended limits after launch. [2]

The same guidance says initial setup happens through the desktop app or desktop web. Mobile conversation access follows where available, but initial creation is not currently supported on mobile. These access details can change as rollout continues. [2]

Ongoing work still has boundaries

OpenAI’s safety explanation says users choose connected accounts and define what the agent may do. Some actions require approval, and some sensitive steps remain with the person. The company also warns that agents can make consequential mistakes. [3]

The practical distinction is between continuing a task and having unrestricted authority to complete every step. A useful assessment therefore depends on the actual work returned, the access granted and whether important decisions were correctly brought back to the user. [3]