Product Launch2 min read

ChatGPT Dots Are Metered by the Work They Do, Not by Chat

By , Senior AI ConsultantPublished

ChatGPT's always-on Dots are free to talk to but count the work they do against your plan, so what you pay depends on the standing jobs you give them.

OpenAI's Dots, the always-on agents it announced in September, are now reaching accounts. Each Dot has its own cloud computer, remembers what you have told it, and keeps working between your conversations. The first one comes with the plan you already pay for, so the real question is what it uses up.

Who can have one now

A Dot comes with the Pro plan, which starts at $100 a month, and with Business Premium in every region where ChatGPT is sold. Enterprise workspaces get it only when their administrator switches it on. Pro subscribers in the UK, Switzerland and the European Economic Area cannot get it yet, and neither can anyone on the free, Go or Plus plans. You create a Dot in the desktop app, and after that you can talk to it from your phone.

A message count cannot measure a Dot

In July 2025, ChatGPT's earlier agent was limited by counting messages: 400 a month on Pro and 40 on Plus. Every run used one, including runs you had scheduled. That was fair, because you decided when each job happened, so your own requests were the whole story.

Scheduling a task is also not what is new. ChatGPT has run scheduled tasks since January 2025. What is new is that a Dot can look through the apps you connect and start research on its own, so some of its work is never a request from you. A limit that counts your requests cannot see that work, so the limit has to measure work done instead.

That is what OpenAI has done. Conversations with a Dot are free. Jobs it hands to ChatGPT Work or Codex count as usual. Its own deeper work comes out of an allowance whose size has not been published. I expect every always-on agent to be priced this way, because once software decides when to act, the person paying no longer controls how much happens.

What to do before the launch period ends

For the first month, a Dot's work comes with extended limits, so the meter is loose exactly while you are learning what to give it. That period ends in late October.

Picture a shop owner on the $100 plan who asks her Dot to read supplier emails every morning, compare new prices with her stock sheet, and draft reorder notes. If that saves twenty minutes each working day, she gets about seven hours a month back, so the plan costs her about $14 for each hour saved. That looks cheap until she adds a fifth and a sixth standing order and has no idea which one uses the most.

The controls help here. In the Dot's custom rules she can choose, for each kind of action, whether it goes ahead without asking, only if she pre-approved it, or only after she approves it. A rule that makes it ask before sending anything keeps her in charge of the part that matters.

A few weeks in, open the Dot's profile in the desktop app and read the Scheduled list. Keep the tasks whose results you actually read, and remove the rest before the launch period ends and the meter starts counting.

Share this

STAY INFORMED

Get AI intelligence like this delivered to your inbox.

Free forever · Unsubscribe anytime


You May Also Find Valuable