Spaces and Dots: OpenAI brings autonomous agents into office work

OpenAI has unveiled two new features in ChatGPT that point in the same direction: getting artificial intelligence agents to work more autonomously, alongside humans. On one side is Spaces, a shared work environment; on the other, Dots, “always-on” agents that pursue a task in the background. The information below comes from OpenAI’s announcement as reported by TechRadar Pro (France) in two articles published on 1 and 2 October 2026. It reflects the company’s own presentation, not an independent assessment of the tools.

Spaces: a common space for documents, colleagues and agents

According to TechRadar Pro, OpenAI defines Spaces as the home for pages and files created, uploaded and shared in ChatGPT, including spreadsheets and presentations. Spaces replaces the former Library feature for Pro, Business and Enterprise users, on the desktop, mobile and web versions of ChatGPT. A single Space can bring together several colleagues as well as ChatGPT, Codex and the new Dots agents.

Spaces also brings Pages, a new document format that combines text, images, charts and interactive tools. What sets it apart is that it is dynamic: instead of a static document that would have to be edited by hand or regenerated by the AI, a Page can stay live and rely on connections to applications to keep its information up to date. Data from Slack, emails, a calendar or other connected services can thus feed its content.

A plugin dedicated to meetings is also entering beta for Pro and Business users on desktop. It can create private, personalised notes, then generate the corresponding follow-up actions directly in a shared Space. TechRadar Pro notes that Spaces is not a direct replacement for full collaboration suites such as Google Workspace or Microsoft 365.

Dots: agents that move forward on their own between conversations

Dots, described by TechRadar Pro as powered by GPT-6, changes the logic of use. Instead of waiting for a succession of prompts, the agent is designed to carry out professional tasks over time. It keeps working between conversations and only comes back to the user when an action requires their attention or authorisation.

The method described by OpenAI is as follows: the user essentially sets goals and limits, then the Dot works out for itself the steps needed to make progress. Each Dot has its own cloud computer and its own browser. It can therefore interact with the web, use applications, run code and work with other plugins, without these operations running directly on the user’s personal computer. Its activity can be displayed as in progress, scheduled or completed.

The examples given stay close to everyday office life: checking a calendar to prepare a report every morning, or analysing information from connected services and applications. According to the source, OpenAI has already integrated Dots into Slack and Microsoft Teams; support for SMS is due to arrive later.

Who has access, and on what terms

The two features are not aimed at the same audience. Spaces is for Pro, Business and Enterprise subscribers. Dots is being rolled out first to eligible users on the Pro and Business Premium plans, at no extra cost for the first Dot. Only the work carried out to reach a goal counts towards usage limits; conversations with a Dot do not consume these quotas.

A Dot’s initial setup must be done from the ChatGPT desktop app or from a browser on a computer; the Dot then becomes accessible from the mobile app. No detailed timetable is given for a wider rollout: OpenAI says it wants to extend the concept gradually as early real-world uses and their cost become better understood. The sources do not specify whether these features are available in Switzerland, nor the price of additional Dots.

What is established, what remains to be verified

Three levels need to be distinguished. Established, according to the source: the launch of the two features, the plans concerned, the isolation of each Dot in a cloud computer and browser, and the integrations with Slack and Microsoft Teams. Announced but not yet available: SMS support and the broader rollout, with no date. Not documented in the sources: how reliable the agents are on real tasks, the details of their safeguards and their cost in use.

On security, OpenAI says that built-in mechanisms are meant to pause or stop a Dot’s work when potentially malicious behaviour or instructions are detected. The sources do not say how this detection works or how effective it is. Likewise, the automatic updating of Pages depends on the quality of the data feeding them, a point the articles do not address.

The questions for Swiss companies

For a Swiss company considering these tools, the announcement raises more questions than it answers. First, on data: where is the information a Dot consults in a calendar, an inbox or Slack processed, and where is the content of a Space stored? The sources do not say.

Next, on control: what limits should be set for an agent that determines its own steps, what access should it be given, and at what point should human approval be required? The principle of a Dot that asks the user about actions requiring their authorisation is described, but not the list of those actions.

Finally, on accountability: if an agent acts on the web, runs code or produces an erroneous report based on connected data, who is answerable for it within the organisation? And how can what it did be documented after the fact? These questions are not answered in the available articles. They deserve to be asked before any deployment, especially as availability in Switzerland has not been specified.

Sources

TechRadar Pro (France), “OpenAI launches Dots, always-on AI agents capable of working on their own in the background” (French-language article; title translated), 1 October 2026.

TechRadar Pro (France), “OpenAI launches ChatGPT Spaces to bring humans and AI agents together in a single workspace” (French-language article; title translated), 2 October 2026.