Home
›
AI news
›
Platform Updates
›
OpenAI's dots put an always-on agent in ChatGPT that keeps working between your conversations
Platform Updates
Confirmed
News

OpenAI's dots put an always-on agent in ChatGPT that keeps working between your conversations

Pro and Business Premium users get one dot included. It runs on its own cloud computer, connects to plugins, and keeps making progress while you are elsewhere.

The Promptifi Team
·
September 29, 2026
·
·
OpenAI
Prompts in the library for this workflow
Browse the library →
Why this matters for sellers

Every AI tool a seller has used so far waits to be asked. A dot does not. It holds a standing brief and keeps working between the moments you open ChatGPT, which changes what the underlying skill is. The rep who gets value here is not the one with the cleverest single prompt but the one who can write a brief a stranger could follow and then check work nobody watched happen. It is also the first agent of this kind included in a plan an individual seller can buy without asking an admin, so the constraint is no longer access. It is whether you can describe your own job clearly enough to hand a piece of it over.

In plain English

"Always-on" means the dot keeps going after you close the chat instead of only replying while you type. "Its own cloud computer" means it has a machine and a browser of its own on OpenAI's side, so it is not driving your laptop. "Plugins" are the connectors that let it reach other apps, and OpenAI says there are more than 4,000. "Read-only tools" during background work means it can look things up but cannot send, post or edit anything until you are in the conversation with it. "Primary dot" is the one included with your plan; more are promised later.

The short version: you get one assistant that keeps working on a job you gave it, it can read on its own but not send on its own, and checking what it produced is still your job.

OpenAI introduced dots at DevDay on September 29, 2026: always-on agents that live inside ChatGPT, work on their own cloud computer with browser access, and keep making progress on a job between your conversations with them. OpenAI calls them "remarkably capable, always-on agents" running on GPT-6 Astra, able to reach more than 4,000 apps through plugins. You start with one primary dot, give it a name, and let it introduce itself.

Announced: September 29, 2026, at OpenAI DevDay.

Availability: Rolling out from September 29 to Pro and Business Premium users, set up in the ChatGPT desktop app or browser. Enterprise, Edu and Healthcare can try the beta when a workspace admin enables it; it is off by default.

Price: Your first dot is included in a Pro or Business Premium plan at no extra cost. OpenAI says you will be able to add more dots later but has not said what those cost.

Regions: "Eligible markets." OpenAI does not list them.

What it doesn't do: A dot does not act freely on your behalf while you are away. During background work it uses read-only tools, so it can research but cannot send messages or change content. Sensitive actions such as changing a password stay with you. OpenAI's safety monitoring can pause or stop a dot's work, and OpenAI states plainly that "dots can still make mistakes, so always review consequential work." The announcement does not name which apps a dot connects to by default, does not state how long one will keep working on a single task, and there is no second dot yet.

What it means for you

SDR — Hand your dot the research task that eats your morning: the accounts landing in your queue tomorrow. A standing brief beats any single prompt here, because it runs whether or not you remember to ask. Read-only background work suits this well, since research is exactly what a dot is allowed to do unsupervised.

AE — The gain is not more research, it is continuity across a long deal. Point the dot at the accounts in your current quarter and ask it to keep a running picture: what changed, who moved, what the last call left open. Then treat the output as a draft. OpenAI says to review consequential work, and on a deal with your name on it, all of it is consequential.

Manager — Your risk is not that reps under-use this. It is a rep handing over judgment work and forwarding output they never read. Before dots spread through the team, decide out loud which tasks are delegable — research, first drafts, summaries — and which are not, meaning anything a customer sees and anything entered in the CRM as fact. Coach the review step rather than the prompt.

RevOps — Two things to settle now. First, this arrives on Pro and Business Premium without you, so assume it is already in your org and find out which plugins reps have connected. Second, the Enterprise beta is off by default and needs a workspace admin, which means you control the enterprise rollout but not the individual one. Write the standard for what a dot may connect to before someone connects one to your CRM.

What to run

The skill a dot rewards is writing a standing brief and then reviewing work you did not watch happen. Our scheduled prompts are the closest starting point — recurring tasks written to run on a cadence rather than on a nudge, which is the same discipline a dot's standing brief demands.

Copy the prompt →
Free · No account needed

The brief for people who carry a number

Two minutes, once a week. What changed in AI, and what to run because of it.

One email a week · No product pitches · Unsubscribe any time
P

The Promptifi Team

Promptifi's AI desk is written and fact-checked by people who carry a number. Drafting is AI-assisted; every item is checked against its primary source by a human before it publishes.