Technology · NEWS EXPLAINER
OpenAI Dots launches always-on AI agents for Slack, Teams and 4,000+ apps
OpenAI’s new Dots can keep projects moving across connected apps and cloud computers, but consequential actions still depend on user-approved rules, permissions or explicit consent.
OpenAI launched Dots at its September 29 DevDay as always-on AI agents designed to keep longer-running projects moving across connected apps, messaging services and a cloud computer. OpenAI says Dots use GPT-6 Astra and can be reached through ChatGPT, Slack and Microsoft Teams.
The launch does not mean every Dot can act without limits. OpenAI’s own documentation says unattended research uses read-only connected-app tools, while consequential actions can require explicit approval or custom rules; workspace owners can separately control app, messaging, local-computer and cloud-browser access.
What OpenAI Dots are designed to do
Dots are positioned as persistent project agents rather than one-shot chat replies. Reuters, The Verge and WIRED describe them as agents that can keep working across multiple steps, revisit evolving tasks and use a cloud computer instead of depending only on a single chat window.
OpenAI says each Dot can connect through plugins to more than 4,000 apps and can communicate through ChatGPT, Slack and Microsoft Teams. That makes the product relevant to workflows that span documents, messaging, research and browser-based tasks rather than a single app.
Always-on does not mean unrestricted
OpenAI’s launch page says proactive unattended research uses read-only connected-app tools. For actions with bigger consequences, the company describes approvals and custom rules that can determine what the agent may do without asking again.
The Help Center adds an administrative layer for workspaces: Enterprise Dots beta access is off by default, and workspace owners can control Dots availability, Slack or Teams access, local-computer access and custom rules. Connected services still apply their own authorization limits.
Who can use Dots now
OpenAI says rollout is beginning for Pro, Business Premium and Enterprise users in eligible markets. Availability can still differ by workspace policy, account eligibility and the permissions granted to each connected service.
That means a user seeing Dots in ChatGPT should not assume every integration or computer-control feature is enabled. Business and Enterprise deployments can be narrower than the headline feature list because administrators and connected apps can impose additional restrictions.
What remains worth watching after launch
The first launch reports also showed that a polished demo is not the same thing as a finished workflow: Reuters noted glitches during live demonstrations, while OpenAI framed the rollout as something that is beginning rather than universally complete on day one.
For readers deciding whether Dots are useful, the practical questions are therefore specific: which apps a workspace has authorized, whether a Dot is allowed to use a cloud or local computer, which actions require approval, and whether the needed channel is enabled. Those controls matter more than the phrase “always-on” by itself.
