OpenAI – GPT Dots
Every developer and small-business owner knows the feeling. You close your laptop at night with a pile of unfinished things: customer feedback nobody has read, a bug you meant to look at, a pull request that still needs tests. The work doesn't stop just because you did.
That is the exact gap OpenAI is aiming at with GPT Dots, announced during its annual DevDay event. I watched the announcement coverage closely, and this is the one that made me stop and rewind.
A quick honest note: this article is based on OpenAI's announcement and the details shared publicly about it. Dots is only just rolling out, so where I describe how I'd use it, those are practical scenarios, not a claim that I've run it for months. The official page is here: OpenAI – Introducing Dots.
So what is GPT Dots, in plain words?
GPT Dots is a persistent AI agent that has its own computer. The most interesting part is that even when you are not actively chatting with it, it can keep working continuously in the background.
Think about how a normal chatbot works. You type a prompt, then you wait for the response. The conversation is the work. When you leave, everything stops.
Dots flips that. It has its own cloud computer and browser, and with the help of those it connects to thousands of applications via plugins. You give it a goal, and it keeps going, 24 hours a day, while you do something else.
What can you actually ask it to do?
The example given in the announcement is a very practical one. You can instruct Dots to:
- continuously monitor customer feedback
- identify and fix bugs
- test new changes
- prepare pull requests
Dots keeps doing all this work in the background, while you are working on another project.
Here is the part I found most useful: while working on one project, Dots can also handle other projects simultaneously. For every new thing, you simply assign new tasks and projects, and it continues executing them.
If you've ever tried to juggle three half-finished jobs and ended up finishing none, you can see why this is appealing.
Where you can use it
You can use Dots across ChatGPT, Slack, and Microsoft Teams.
That matters more than it sounds. Most of us don't want to open yet another app. If your team already talks in Slack or Teams, you can hand a task over in the same place you'd message a colleague. And because it works from your phone or any other device, you can assign something from the bus and check on it later.
Who gets it, and where?
This is where people will get frustrated, so read carefully:
- Dots is rolling out for ChatGPT Pro and Business Premium users.
- It is available in eligible markets, which currently excludes the European Union.
- If you are on an eligible plan, you'll find the Dots option in the left-hand sidebar menu of ChatGPT.
If you don't see it yet, don't panic. Rollouts take time, and your plan and region both matter.
Is this a new idea?
Not entirely. Competitor systems like Meta Muse and Grok's GrokBot are working in similar directions. The difference, according to the announcement, is that OpenAI has packaged this entire experience into an easy-to-use consumer product.
That "easy to use" bit is the real story. Always-on agents have been possible for people willing to wire together scripts, servers, and APIs. Putting one inside a sidebar button is a different level of access.
How I'd get started (a step-by-step approach)
Here's the approach I'd take. It's sensible with any autonomous tool, not just this one.
Step 1: Check your plan and region. Make sure you're on Pro or Business Premium and in an eligible market. Look for Dots in the left sidebar of ChatGPT.
Step 2: Start with one small, low-risk job. Don't hand it your entire company on day one. Something like "monitor customer feedback and summarize themes" is a safe first task, because reading and summarizing is hard to get badly wrong.
Step 3: Write the goal like you'd brief a new teammate. Say what "done" looks like, how often you want updates, and what it should not touch. Vague goals produce vague results, and with something running 24/7, vague results pile up fast.
Step 4: Connect only the plugins you need. Thousands of integrations sounds exciting, but every connection is access to something real. Start narrow.
Step 5: Check in, then widen the scope. Once the first task behaves well, add the next one. Dots can handle multiple projects at once, but you'll trust it more if you build up gradually.
A realistic scenario
Say you run a small software product with a two-person team. Customers send feedback through email and a support form, and it ends up ignored until Friday.
With a tool like Dots, you could assign it to watch that feedback, spot repeated complaints, look into likely bugs, test changes, and prepare a pull request for a human to review. That's the flow OpenAI described, and it's easy to picture how it saves the "I'll get to it later" backlog.
The key phrase there is for a human to review. I'd keep it that way. Let it prepare, and you approve. You can read how pull requests work in GitHub's documentation if you want to set up a review habit before handing anything over.
Common mistakes to avoid
Giving too much access too soon. An agent with its own computer and browser, plus plugin access to your apps, can do a lot. That's the point, and also the risk. Grant what's needed, nothing more.
Skipping the review step. Autonomous doesn't mean infallible. Always read what it produces before it reaches customers or your main codebase.
Writing fuzzy instructions. "Make my product better" isn't a task. "Review this week's feedback and list the top three recurring bugs" is.
Assuming it's available everywhere. If you're in the European Union, it's not available right now. Don't pay for a plan upgrade expecting Dots if your market isn't eligible. Check first.
Treating it like a chatbot. The whole shift is that you don't have to sit and wait. Assign, step away, and check back.
Why this matters
For years, AI tools have felt like very smart assistants who only work when you're standing over them. Dots is an attempt at something closer to a colleague who keeps working after you log off.
Whether it lives up to the promise will depend on how reliable it is in real daily use, and that's something only time and plenty of users will show. But the direction is clear, and the combination of a persistent agent, its own cloud computer, thousands of plugins, and access through ChatGPT, Slack, and Microsoft Teams makes this one of the more interesting announcements from DevDay.
If you're on an eligible plan, open your ChatGPT sidebar and look for Dots. Start small and see what happens.
Related reading on this site (update these links to match your own posts):
Frequently Asked Questions (FAQs)
What is GPT Dots?
GPT Dots is a persistent AI agent from OpenAI that has its own cloud computer and browser. It can keep working on assigned goals in the background, even when you're not actively chatting with it.
How is GPT Dots different from a regular chatbot?
A regular chatbot waits for your prompt and replies. Dots runs continuously, 24/7, working toward goals you've assigned without needing you to prompt it each time.
What can GPT Dots do?
According to the announcement, you can instruct it to monitor customer feedback, identify and fix bugs, test new changes, and prepare pull requests. It connects to thousands of applications through plugins.
Can GPT Dots work on more than one project at once?
Yes. While it works on one project, it can handle other projects simultaneously, and you can keep assigning new tasks.
Where can I use GPT Dots?
You can use it across ChatGPT, Slack, and Microsoft Teams, which means you can interact with it from your phone or any other device.
Who can use GPT Dots?
It's rolling out to ChatGPT Pro and Business Premium users in eligible markets. It is currently not available in the European Union.
Where do I find GPT Dots in ChatGPT?
If you're on an eligible plan, look for the Dots option in the left-hand sidebar menu of ChatGPT.
Is GPT Dots the only always-on AI agent?
No. Competing systems such as Meta Muse and Grok's GrokBot are moving in similar directions. The difference highlighted is that OpenAI has packaged the experience into an easy-to-use consumer product.
Is it safe to give an always-on agent access to my apps?
Treat it like any new team member: give access only to what's needed, start with low-risk tasks, and review its work before anything goes live.
Where can I read the official announcement?
On OpenAI's page: Introducing Dots.
