7 min read
A colleague who works while you sleep. In Italy, not for everyone (yet)
OpenAI has launched dots, agents that work inside ChatGPT day and night. What they do, why they don't reach everyone in Italy and what's worth preparing in the meantime.
In short
Dots are ChatGPT agents that work even when you don't ask them anything. The work that makes them useful can already be done today: data in order and clear rules.
- Name
- dots
- Made by
- OpenAI
- Cost
- Included in the paid Pro and Business Premium plans
- Where
- In the ChatGPT app and inside tools like Slack and Teams
- In Italy
- Pro excluded for now, Business Premium on its way
There's a sentence we often hear from people who run a small business: "I'd need someone to take care of all those little things."
Invoices to send. Emails to answer. The supplier to chase. The document to dig out from among hundreds of attachments. Nothing difficult. But all together, at the end of the day, that's two hours gone.
On 29 September 2026 OpenAI presented something that tries to answer exactly that sentence. They're called dots.
What they are, without beating around the bush
A dot is an artificial intelligence agent. "Agent" means it doesn't just answer when you write to it: it works on its own, even while you're doing something else.
Each dot has its own computer, which isn't at your place but on OpenAI's servers, with its own browser. It runs around the clock, connects to thousands of apps (email, calendar, shared documents), follows several jobs at once and learns from your corrections. You'll find it in the ChatGPT app and also inside tools like Slack and Teams.
- EarlyReminds you of this week's deadlines
- MorningSuggests the reminder to the supplier
- LunchFinds a document among hundreds of attachments
- AfternoonPrepares draft replies
- EveningAsks for your OK before sending
The example that went around the web
One of the first people to try it told a simple story. He's a freelance writer and had forgotten to send an invoice. His dot noticed while reading his emails, took the right details from the conversation, prepared the invoice and proposed it to him. He said OK, and the invoice went out.
Nobody had asked it anything. It just noticed.
That's the difference from what we've used so far: with a chat, you have to remember to ask. With a dot, in theory, it remembers for you.
- EveningYou close your computer
- At nightReads emails (read-only)
- At nightFinds an unsent invoice
- At nightPrepares the draft
- In the morningAsks for your OK
"What if it causes damage?"
It's the first question we asked ourselves too. OpenAI has set some boundaries:
- In the background it can only read. When it works alone it looks and suggests, but doesn't send messages or change anything.
- You decide the rules, action by action.
- Sensitive actions always ask for permission, such as changing a password.
- Its computer is separate from yours, unless you decide to connect them.
- A monitoring system can pause it if it sees something wrong.
And an honest sentence, written by OpenAI itself: agents can still make mistakes, so important work should be double-checked.
The detail few people mention
Dots are included in ChatGPT's most expensive subscriptions, Pro and Business Premium.
According to several outlets, however, for Pro subscribers the European Economic Area, Switzerland and the United Kingdom are excluded. So Italy too. OpenAI's official page only mentions "covered markets", without listing them. Things are a little better for companies: according to OpenAI's help page, on the Business Premium plan dots are arriving here too. If your work depends on this, ask OpenAI to confirm for your account before you plan around it.
Why the difference? OpenAI doesn't explain. It probably has to do with European rules on privacy and artificial intelligence, which are stricter than American ones. It's not the first new feature to reach Europe later, and it won't be the last.
What to prepare now
An agent, however smart, is only as useful as the information it finds and the rules you give it. If your business is in a mess, you'll get an agent that works very fast in the mess.
Three things to do now, with or without dots.
1. Line up your information. Clients, orders, quotes, invoices: where are they? If the answer is "a bit in emails, a bit in a spreadsheet, a bit in Giorgio's head", no agent will read Giorgio's head. Gathering everything in a few places makes your life simpler from tomorrow.
2. Set the boundaries. Take a sheet of paper and make three columns:
| Can do it alone | Must ask you first | Must never do it |
|---|---|---|
| Prepare drafts | Send an email to a client | Payments |
| Search for documents | Confirm an order | Sign contracts |
| Remind you of deadlines | Move an appointment | Touch sensitive data |
3. Think about your clients' data. An agent connected to email and management software sees everything you see. Before connecting it, talk to whoever handles privacy in your company. It's not bureaucracy: it's respect for the people who trust you.
In practice, for your business
- Gather clients, orders, quotes and invoices in a few places.
- Make the three-column sheet: what it can do alone, what it must ask you, what it must never do.
- Before connecting an agent to email and management software, talk to whoever handles privacy in your company.
Our detour
No rush. People selling the "agent revolution" as something to grab right now, or be left behind, usually have something to sell you.
Agents will come, from OpenAI or someone else. And the work that makes them useful isn't technological: it's putting things in order, deciding who does what, protecting data. It's done with a sheet of paper, a coffee and a bit of patience.
Those who do it now will switch agents on in a week when they arrive. Those who don't will take six months.
If you'd like a hand figuring out where to start, write to us. We'll do the three-column sheet together.
Sources
Accessed on 7 October 2026