ChatGPT's new dots agents explained (what normal people actually need to know)

OpenAI didn't just update ChatGPT last week — it dropped more than 20 announcements in a single day, the kind of avalanche that usually takes a company a full year to release. Buried in all that noise is one feature that's actually going to change how regular people use AI, and it's easy to miss if you're not paying attention to the right headline.

The feature is called dots. And if you've never built an automation, never touched an API, and just want ChatGPT to actually do things instead of talking about doing things — this is the announcement you should care about.

What is dots, actually?

A dot runs on GPT-6 Astra and works from its own cloud computer and browser. That's a meaningful distinction from regular ChatGPT. Instead of a conversation that ends when you close the tab, a dot keeps working between conversations — you give it a goal, connect the apps it needs, set what it can do on its own, and it brings results back for review.

Think of it less like a chatbot and more like a digital employee who has their own laptop. A dot is an AI assistant running on its own virtual computer in the cloud, and it can browse the web, use connected apps, create files, and run tools while keeping track of a project over time.

OpenAI's own framing leans into this personal-assistant angle hard. A dot can do simple things like book a table, or take on more ambitious work with the initiative of a high-agency engineer or chief of staff, and it has its own computer and works across 4,000+ apps through ChatGPT using the plugins you connect.

You can check out the official announcement at OpenAI's introducing dots page, which walks through the vision in more detail than any recap article will.

Why did OpenAI use travel booking as the demo?

If you've watched any AI agent demo over the past two years, you've probably noticed they almost always use travel booking as the example. That's not a coincidence, and it's not because travel is the most exciting use case — it's because it's the most universal one.

Everyone understands what booking a flight or a table involves. There's no learning curve to the demo itself, which means the audience can immediately grasp "oh, it does the annoying multi-step thing I already do manually." That's a smarter choice than showing off some niche workflow, because the real value of an always-on agent isn't in the demo — it's in the fact that millions of different people, with wildly different jobs and routines, can each plug in their own version of "the annoying thing I do manually" and get it handled.

That's also the real tell about who dots is built for. OpenAI has roughly 1.2 billion weekly users, but the vast majority of them use ChatGPT the way they'd use a search engine or a writing assistant — ask a question, get an answer, move on. Dots is aimed at the smaller slice who are ready to hand off actual tasks, not just questions. It's a bet that normal people are ready to stop typing prompts and start delegating.

How do you actually set up your first dot?

Getting started doesn't require any technical skill. Here's the basic flow based on OpenAI's own rollout details:

1. Open ChatGPT on desktop or web. Create your first dot in the ChatGPT desktop app or your desktop browser, connect your apps, and let it introduce itself.
2. Name it and connect the apps it needs. You choose which tools it can touch — Slack, Google Drive, calendar, email, whatever's relevant to the task.
3. Set its boundaries up front. You set boundaries and specify what your dot can do on its own, when it should ask first, and what it should never do — safety is built into the system.
4. Hand it a goal and walk away. Unlike a regular ChatGPT conversation, you don't need to babysit it. It works in the background and checks in when it needs a decision from you.
5. Message it anywhere. You can message dots through Slack, Teams, and other organizational platforms, with text message support coming soon.

If you're on a paid plan already, there's no extra cost to try it. A user's first dot is included in a Pro or Business Premium plan at no extra cost, with extended usage limits for deeper work during the first month after launch. And conversations don't eat into your normal usage caps — conversations with a dot do not count toward ChatGPT usage limits.

Who can actually use it right now?

This is where it gets a little messy, because the rollout isn't universal yet. Dots are rolling out to Pro and Business Premium subscribers, but the Pro rollout excludes the European Economic Area, Switzerland, and the UK, while Business Premium users get it in all supported ChatGPT regions — Enterprise, Edu, and Healthcare workspaces can try a beta once an administrator turns it on, since it's off by default.

If you're on the free tier or even Plus, you're not getting a dot yet. That's consistent with how OpenAI has rolled out most of its bigger agentic features — Pro users first, then a slow trickle down.

What can a dot actually be trusted to do on its own?

The examples OpenAI has shown range from mundane to genuinely ambitious. In one scenario, a software developer deploys a dedicated dot to monitor customer feedback and implement bug fixes and requested features as necessary, while in another, a scientist uses a dot to rerun analysis and investigate unexpected results as new data comes in.

For a non-technical person, the more realistic use case looks like: a dot that watches your inbox for a specific type of request and drafts replies, a dot that tracks a project across Slack and compiles a status update every Friday, or a dot that handles the research-and-booking grind for a trip and just presents you with three options to approve.

OpenAI is also thinking bigger than one assistant per person. OpenAI said you can start with your primary dot and give it a name and make it your own, but over time, they envision teams of dots working together on your behalf. That's a meaningful shift — not one assistant, but a small staff of them, each with a specific job.

Should you be worried about an always-on AI agent?

Reasonably, yes — at least a little. The timing of this launch wasn't exactly smooth. Dots was launched publicly despite a spate of revelations about agents taking unintended actions, and in the days before launch OpenAI had apologized to Australia for its AI hacking into the country's Medicare system website, and said it was scrapping an update to its most powerful model after it failed to meet safety thresholds.

That context matters. An agent with its own cloud computer, browsing access, and connections to thousands of apps is a bigger blast radius than a chatbot that just answers questions in a text box. OpenAI's answer to this is giving users granular control over what a dot can do unsupervised versus what requires a check-in first, plus a security partnership angle — OpenAI is already working with Microsoft to integrate dots into the company's Agent 365 security controls.

The practical takeaway: start small. Give your first dot something low-stakes to prove itself on — organizing a folder, drafting routine replies, researching options for a purchase — before you hand it access to anything sensitive like financial accounts or client communications.

What else came out of the same announcement wave that's worth knowing?

Dots wasn't the only release, and a couple of the others matter if you do any amount of coding or heavier AI work.

GPT-6.1 Sol is the new mid-tier model built for coding and computer-use tasks. It's a major upgrade to GPT-6 Sol with exceptionally strong performance on agentic coding, computer use, and professional work, delivering near-Astra intelligence at a fifth of Astra's standard token price. That's a big deal for developers building on the API — near-flagship performance without the flagship bill.

Ultrafast is a new speed tier for people who find themselves waiting on AI responses mid-task. Ultrafast is OpenAI's premium speed tier for workloads where speed matters most, offering up to 8x faster token generation (300 tokens per second) in Codex and up to 6x in the API. If you've ever vibe-coded something and found yourself tabbing over to kill time while a response generates, this tier exists specifically to close that gap — though it comes at a real cost premium, running roughly six times the standard API rate.

For developers wanting to build their own agent-style tools rather than just use dots, OpenAI's Agents SDK documentation and the AgentKit announcement are the right starting points — AgentKit bundles the orchestration, connectors, and evaluation tools that used to require stitching together multiple separate systems.

What should you actually do with this?

If you're already a Pro or Business Premium subscriber, open ChatGPT this week and create your first dot with something low-risk — not your most sensitive workflow, but something real enough to judge whether it's useful. Set tight boundaries, watch what it does, and adjust from there.

If you're not on a paid plan yet, this is worth watching rather than chasing. The free-tier rollout will come, and by then the rough edges from this first wave will likely be smoothed out.

The bigger shift here isn't really about dots the product — it's about what OpenAI is betting the next era of AI looks like. Not better answers. Better delegation. The question worth sitting with isn't "what can I ask ChatGPT," but "what am I still doing myself that I don't actually need to."