By Phil Gee · AI Academy, Southport
On 29 September 2026, OpenAI announced Dots at its developer event in San Francisco. If you use ChatGPT, this is the biggest change to it in a while, so let's break it down without the jargon.
What is a Dot?
A Dot is an AI agent that lives inside ChatGPT. You give it a name, a job and some boundaries, and it carries on working after you've closed the chat. That's the headline: it doesn't wait for you to type something.
Each Dot has its own cloud computer and its own web browser, and it connects to your other apps through plugins. OpenAI says it can reach more than 4,000 apps, including tools like Slack and Microsoft Teams, so you can talk to it from there as well as from ChatGPT.
One of the examples OpenAI shared was a Dot that noticed an invoice had never been sent, prepared it, and sent it once the owner approved. Small thing. Very useful thing.
It's proactive, but on a lead
When it's idle, a Dot can do what OpenAI calls proactive research, looking into things it thinks you'll need. In that mode it's read-only, so it can't send messages or change anything. For actions that actually do something, you set approval rules, and OpenAI is clear that you should always review important work.
That's sensible. An agent that can send emails and spend your money needs a firm hand on the lead.
Who can use it?
This is where most people will be disappointed:
- It's for paying customers at the top end. At launch, Dots are available to ChatGPT Pro and Business Premium subscribers in eligible markets. Free, Go and Plus users don't get them yet, and OpenAI hasn't given a date.
- The UK may be left out for now. Some reports say Pro users in the UK, EU and Switzerland won't have access at launch. Check OpenAI's own page for the current position before planning anything.
- It's not cheap. One report puts the entry point at around 100 US dollars a month, and OpenAI hasn't yet said what extra Dots or heavier use will cost.
What should you expect?
Expect big promises and a bumpy start. Agents that run on their own are a genuine step forward, but they're also where the mistakes get more expensive, because an AI that acts can get things wrong in the real world, not just on a screen.
My advice is simple. If you do get access, pick one well-defined job and try it as a pilot. Something like chasing unpaid invoices or pulling together a weekly report. Watch how it behaves for a few weeks before you give it anything more important.
If you can't get Dots yet, don't worry. You haven't missed the boat. The skills that make agents work are the ones I teach every day: knowing how to describe a job clearly, giving the AI context about your business, and checking its work. Learn those and you'll be ready for Dots, Muse or whatever turns up next month.
Want to get those foundations in place? Book a free chat with me and let's talk about where AI could genuinely save you time.