Dots is OpenAI’s new AI assistant, unveiled at DevDay on September 29, 2026, and powered by the GPT-6 Astra model. Unlike ChatGPT, Dots isn’t tied to a single chat window—it runs continuously in the background toward goals the user sets, and it plugs directly into the tools teams already use, like Slack and Teams.
But convenience comes at the price of trust. Users should know what data Dots can see, how much it can do on their behalf, and what the total cost really is. If privacy controls stay unclear or it slips up too often, the cute avatar won’t be enough to make it useful in practice.
What Dots Is and Where It Came From
OpenAI introduced Dots at DevDay, describing it as an “agent that works around the clock and can be handed almost anything.” Under the hood is the GPT-6 Astra model. Unlike ChatGPT, which is anchored to a chat window, Dots runs as an independent entity pursuing a goal continuously. Users can name their primary Dot, and OpenAI says a future version will let “teams of Dots” coordinate on a person’s behalf.
The Features Dots Actually Ships With
The first standout is that it plugs straight into tools teams already rely on, namely Slack and Teams, with SMS support coming later. Organizations can also spin up “specialist Dots,” each responsible for a different slice of work, provisioned with the credentials and tools they already have—no need to rebuild access from scratch.
OpenAI’s own examples: a software team assigns a Dot to track customer feedback and propose bug fixes or new features, while a scientist uses one to automatically rerun analysis whenever new experimental data comes in, without triggering it manually each time.
How Dots Differs From Regular ChatGPT and ChatGPT Agent
| Factor | Legacy ChatGPT | Dots |
|---|---|---|
| Task initiation | Waits for the user to give an instruction | Starts from a goal set in advance |
| Continuity of context | Context must be re-entered in a new chat | Keeps tracking the same goal in the background |
| Where it runs | Tied to the chat window | Runs through Slack, Teams, and ChatGPT |
| Tool usage | Used only when requested | Chooses tools on its own based on the goal |
| Level of oversight | The user controls every step | The user sets boundaries and reviews results later |
The difference is that Dots is built for work that needs following up inside the tools a team already uses, like chasing feedback in Slack or rerunning an analysis. But the more autonomy it gets, the more clearly users need to define where approval is required.
How Similar Are Dots and Meta’s Muse
Both companies took the same route: making an AI agent look like a cute cartoon character instead of a blank text box. Dots and Muse share a strikingly similar rounded, brightly colored persona, which is why the launch reads as a direct jab at each other in the personal-AI-assistant market. The functional differences between the two are still something to watch as real-world use grows.
| Factor | Dots | Meta Muse | ChatGPT Agent |
|---|---|---|---|
| Avatar persona | Round, brightly colored | Round, brightly colored—very similar | No avatar, works behind the scenes |
| Enterprise connections | Slack, Teams | Focused on Meta's own services | Through ChatGPT and developer tools |
| Main pitch | Works continuously on a team's behalf | An approachable personal assistant | Completes multistep tasks toward a goal |
Who Gets Access First, and Under What Terms
Dots became available on September 29, 2026, for ChatGPT Pro and Business Premium users in a limited set of eligible markets, accessible through ChatGPT itself or through Codex. OpenAI also said it’s working with Microsoft to integrate Dots with Agent 365’s security controls—a signal that enterprise customers will get more permission-management tooling down the line.
What Dots Gets Right, and What Still Needs Trust
Dots looks friendly and approachable thanks to its lively avatar, and it cuts busywork by operating continuously inside tools a team already uses. But users still need to know what it can access behind the scenes, and should review its output before letting it handle anything important.
Pros
- +Plugs into Slack and Teams right away, no need to switch tools
- +Specialist Dots can be assigned to different tasks
- +Handles ongoing work that would otherwise require constant follow-up
- +Friendly persona lowers the barrier to getting started
Cons
- −Requires fairly broad access to internal systems and data
- −Limited to Pro and Business Premium plans in select markets for now
- −No separate, detailed pricing has been published yet
- −Still unclear how far Agent 365 oversight will actually reach
The Cost That Doesn’t Show Up on the Pricing Page
A Pro or Business Premium subscription is only the cost you can see. A Dot running continuously in the background still consumes quota and compute every time it acts, and the more it connects to Slack, Teams, or other systems, the more someone needs to keep reviewing its access and its output.
The cost that doesn’t show up on the pricing page is the risk of letting a Dot reach into customer data, internal feedback, or code under active development. Without a clear approval step, the convenience can end up costing time spent cleaning up after it later.
How Much Authority Should Dots Actually Get
The real question isn’t how capable Dots is, but how an organization decides to bound its authority. Repetitive, reversible work—like summarizing feedback or rerunning an analysis—can reasonably be left to a Dot.
Anything touching money, customer data, or code headed to production should require human approval first, and work that risks privacy or causes damage that can’t be undone shouldn’t be delegated to it at all.