OpenAI Dots, Explained: A Beginner's Guide (2026)

Ellie at Willa, Editorial lead, Willa · Published September 30, 2026 · Updated September 30, 2026

On Tuesday, OpenAI used its DevDay stage to launch Dots, which the company describes as "remarkably capable, always-on agents built to handle everything." If that sentence made you tired instead of excited, this guide is for you. Most of the coverage so far is written for developers and investors. Here is what Dots actually are, who can use them, and what they mean for you if you work in marketing, ops, recruiting, program management, or any other job where nobody pays you to read API docs.

What are OpenAI Dots?

A Dot is an AI agent that keeps working after you close the app. Each one runs on its own cloud computer with its own browser, powered by OpenAI's GPT-6 Astra model, so it can research, draft, and organize in the background while you do something else entirely. That is the real difference from the ChatGPT you already know. A chat ends when you stop typing. A Dot keeps going.

You talk to your Dot the way you talk to a colleague: message it or call it through ChatGPT on desktop, web, or mobile, or reach it in Slack or Microsoft Teams, with texting support promised soon. It connects to more than 4,000 apps through plugins, learns your preferences from feedback, and can juggle several projects at once. OpenAI even encourages you to name your primary Dot and "make it your own."

One early example that stuck with us, from The Next Web's coverage: a freelance writer's Dot noticed an invoice he needed to send, pulled the details from the email thread, and drafted the invoice before he asked. That is the pitch in miniature: a helper that notices work needs doing before you ask.

Who can get Dots, and what do they cost?

Dots are included at no extra cost with ChatGPT Pro and Business Premium plans. You get one Dot per subscription for now, with the option to add more coming later, according to 9to5Google.

The catches worth knowing before you get excited:

• Free and Plus subscribers do not get Dots yet. This is a Pro and Business Premium feature at launch.
• Pro users in the European Economic Area, Switzerland, and the UK do not have access yet, though Business Premium users do in all supported ChatGPT regions.
• Enterprise, Edu, and Healthcare accounts get beta access only if an admin turns it on, so if you are on a company plan, that conversation goes through IT first.

One pleasant surprise on limits: conversations with your Dot do not count toward your ChatGPT usage limits, though tasks you kick off in Codex or ChatGPT Work still do.

What can a Dot actually do for your job?

Strip away the demo-stage language and the useful pattern is: recurring, background, multi-step work that today lives on a sticky note. Some examples that map to real non-engineering jobs:

• Marketing: monitor mentions and competitor launches overnight, then hand you a digest with drafts of possible responses each morning.
• Operations and admin: watch an inbox for invoices, receipts, or forms, extract the details, and keep a running tracker current.
• Recruiting: research candidates or companies in the background and prepare briefing notes before your calls.
• Program and project management: chase status across the tools your team already uses and flag what slipped, instead of you spending Friday collecting updates.
• Freelancing and solo business: exactly the invoice example above. Follow-ups, drafts, and paperwork that fall through the cracks when you are the whole company.

OpenAI is also piloting "specialist Dots" for enterprise roles like procurement, invoicing, marketing, support, and contracting, and is working with Microsoft to fit Dots into Agent 365 security controls. Translation: your company may eventually hand you a pre-configured Dot for your exact role.

How do you set up your first Dot?

The setup is short, but the order matters. Per 9to5Google's reporting, creation happens in the ChatGPT desktop app, and after that you can work with your Dot from mobile too.

1. Open the ChatGPT desktop app on a Pro or Business Premium account and create your primary Dot. Give it a name you will not be embarrassed to say in a meeting.
2. Connect only the apps your first real task needs. With 4,000 plus plugins available, the temptation is to connect everything. Resist it. Start with one or two.
3. Give it one recurring task you already do manually, described the way you would brief a new assistant: what to watch, what a good result looks like, and when to check with you.
4. Open its cloud computer view and watch it work the first few times. You can see everything the Dot does in real time, which is the fastest way to learn what it is good at.
5. Correct it out loud. Dots learn your preferences from feedback, so treat the first week like onboarding, not like a vending machine.

That first-week investment is the same advice we give in How to Build an AI Teammate That Does Real Work: an agent gets useful when you manage it like a junior teammate, with clear briefs and honest feedback, not one-line commands.

Are Dots safe to connect to your accounts?

OpenAI has clearly thought about the obvious fear, which is an autonomous agent doing something you did not approve. The controls, as reported by The Next Web: Dots run on isolated cloud infrastructure separate from your devices, use saved logins without exposing your passwords to the model, and do background research in read-only mode, meaning they cannot send messages or change content on their own while researching. You can set custom rules to allow, block, or require approval for specific kinds of tasks. Actions that affect your accounts or share information get an automatic review step, and password changes always require your approval. A monitoring system can pause a Dot if something looks wrong.

Two honest cautions. First, OpenAI's own warning: "Dots can still make mistakes, so always review consequential work." Second, agents in general are still new territory. The Next Web notes an earlier incident in which OpenAI agents posted 53 users' images online. So our practical advice is: keep your Dot away from anything you could not recover from (client bank details, sensitive HR files), require approval for anything that leaves your accounts, and if you use it for work, check your company's AI policy first. We wrote about how to handle that conversation in Should You Tell Your Boss You're Using AI at Work?

How do Dots compare to agents you can already build?

If you have followed along with our agent tutorials, Dots will feel familiar. Claude's cloud sessions and scheduled tasks can already do background work on a timer, and plenty of Willa members run agents like that today without writing code. What Dots change is packaging and price. The always-on cloud computer, the app connections, and the persistent memory come pre-assembled and bundled into a plan many people already pay for, with no setup beyond the steps above.

The honest comparison: if you want one assistant that lives inside the OpenAI ecosystem and you have a Pro plan, Dots are the lowest-effort way to get an always-on agent right now. If you want to understand how agents work, control every step, or you use Claude as your main tool, building your own still teaches you more and keeps you flexible. Our guide How to Build Your First AI Agent (No Code Needed) is the place to start on that path, and the concepts transfer directly to managing a Dot.

Also announced at DevDay, for context: a new GPT-6.1 Sol model that OpenAI says offers near-Astra performance at about one fifth the token cost, while the planned October launch of GPT-6.1 Astra was cancelled after failed safety tests, per The Next Web. Worth remembering when the hype gets loud: even the labs are still learning what these systems can safely do.

Key takeaways

• Dots are always-on OpenAI agents with their own cloud computers, launched September 29, 2026, that keep working in the background after you close the app.
• They are included with ChatGPT Pro and Business Premium at no extra cost, one per subscription, but Pro users in the EEA, Switzerland, and the UK are excluded at launch.
• Setup happens in the ChatGPT desktop app: create and name your Dot, connect a couple of apps, and brief it on one recurring task.
• Safety controls are real (read-only research, custom approval rules, isolated infrastructure), but OpenAI itself says to review consequential work, so start with low-stakes tasks.
• The skills that make a Dot useful are the same ones that make any AI agent useful: clear briefs, honest feedback, and knowing which work to hand off.

New tools like this land almost every week now, and figuring them out alone is the slow way. Willa hosts expert-led sessions almost every week for women building careers with AI, including hands-on walkthroughs of exactly this kind of launch. See what's coming up on our events page, or become a member and compare notes with women who set up their first agent last month and can tell you what they wish they had known.

• openai dots
• ai agents
• chatgpt
• devday 2026
• ai skills

Join Willa, the community where women learn AI together.

All posts