A ChatGPT dot is an assistant that keeps working toward a goal between your conversations, on its own cloud computer, with the apps you connect to it. You give it a standing job, and it messages you with progress, questions and decisions. OpenAI launched dots on September 29, 2026.
The jobs worth giving it are the ones you repeat every week. Here are 5, each taken from OpenAI's own examples or from a hands-on test by 2 YouTube creators with early access, Skill Leap AI and Futurepedia. Each job has what it does, why it helps, and 1 thing to try. Where the reasoning is ours, it says "Our read".
Watch the 44-second explainer on YouTube. Jobs 4 and 5 are only in this article.
What a dot is, and who can get one
What it is. OpenAI calls a dot an “always-on agent”. It runs on GPT-6 Astra, with its own cloud computer and browser. It connects to “over 4,000 apps” through plugins, and OpenAI says it “can work towards your goals 24/7”. When you aren't working with it, it “looks for ways to help in the background”. Skill Leap's host sums up the difference from a chat: “it messages you. You don't message it.”
Where you talk to it. In ChatGPT on desktop, web and mobile, in Slack and Teams, and on voice calls. Texting is a limited beta for US Pro users. You create a dot on desktop only. OpenAI's help article, read October 9, 2026, says a dot can't call you and can't have its own email address yet, but you can connect your personal email.
Who can get one.
- ChatGPT Pro users “in markets excluding the European Economic Area, Switzerland, and the UK”.
- Business Premium users in all supported regions.
- Enterprise, Edu and Healthcare, as a beta that stays off until an admin turns it on.
OpenAI is rolling it out gradually, so it may not be on your account the day you look.
What it costs. OpenAI says the first dot is “included in your Pro or Business Premium plan at no extra cost”. Talking to your dot doesn't count toward your ChatGPT usage limits, but the Codex or ChatGPT Work tasks it starts do. OpenAI says you'll be able to add more dots later. The help article says "Pro" without naming a Pro tier, so check your own plan before you count on a dot.
1. Turn 1 transcript into posts and show notes
What it does. This is OpenAI's own example, from its announcement: “When a new interview transcript arrives, it identifies moments for clips, prepares show notes, and drafts social posts for your approval.”
Why it helps. Our read: if you publish interviews, this is the work you redo by hand every episode. Here the posts come back to you as drafts for your approval. It's also the job that needs the least access: 1 folder of transcripts, and nothing that sends, posts or spends.
Try this. "Every morning, check my transcripts folder. For each new one, pick 3 clip moments and draft 3 posts for my approval."
2. Keep launch materials in sync
What it does. Skill Leap's host tested a dot in early access with a folder of launch documents for a demo product. He gave it 3 Custom Rules, in his words: “Draft anything, but never send or post anything. Never change the price and anything that could move the launch date comes to me first.” The dot found “a pricing mismatch between a couple of the different documents”. After about 10 minutes, it had rewritten 3 social post drafts and the launch email. OpenAI names the same kind of work in its announcement: “Revise a launch as scope changes”.
Why it helps. Our read: when a price or a date changes, every asset that mentions it has to change too, and any copy you miss is the copy a customer sees. Checking every document against the others is the step you skip on deadline.
Try this. Give it the folder and the launch date, and start with his first rule, word for word: “Draft anything, but never send or post anything.” Then ask: "Compare every price, date and product name across these documents, and list each mismatch before you change anything."
3. Get a morning brief
What it does. OpenAI's help article: “you could ask it to review your calendar each morning and tell you what’s coming up.”
Why it helps. Our read: with your email connected, the brief can cover the replies you owe, not just your meetings. It also turns a stream of pings into 1 message. Skill Leap's host said that over a weekend his dot “was sending me too many messages”, so he told it: “just give me a recap once a day unless it's urgent.”
Try this. "Each weekday at 8 AM, tell me what's due and what needs my reply. Between briefs, message me only if something is urgent."
4. Always have the next piece ready
What it does. Skill Leap's standing job for his dot: “always have my next piece of content ready before I finish the one I'm working on.” It found his 5 scripts in Google Drive and kept working.
Why it helps. Our read: if it keeps up, your next draft is waiting when you finish the current one, so you don't start from a blank page. A chat answers 1 prompt and stops. A standing job keeps running without you restarting it, which is what Skill Leap means by giving it “a responsibility, not a prompt”.
Try this. Point it at your drafts folder and your content calendar: "Keep my next post drafted before I publish the current one. Follow my content calendar, and ask before you change a date."
5. Run a whole project from 1 chat
What it does. Futurepedia's host used his dot as “the orchestrator” of his ChatGPT chats, ChatGPT Work and Codex threads. In 1 request, it took 3 steps: it wrote 3 video intros in ChatGPT Work, picked the best and turned it into motion graphics in Codex, and started a thumbnail in another chat. He told it: “Bring me decisions when you need me.” What he noticed: “It will prompt itself with a lot of detail that you might not put into your prompt.”
Why it helps. Our read: you delegate from 1 place instead of juggling threads and retyping the context into each one. Futurepedia's verdict is the same: it “worked best for me as a bridge to all my threads and projects”. 1 cost to watch: the ChatGPT Work and Codex tasks a dot starts count toward your usage limits, even though talking to it doesn't.
Try this. Give it 1 project and ask for decisions, not updates: "Take this video from intro script to thumbnail. Ask me only when you need a decision."
1 rule for all 5
Set “Ask before taking action” for anything that sends, posts or spends. In Custom Rules, each action a dot takes can be set to 1 of 4 behaviors: “Take action without asking”, “Take action if pre-approved”, “Ask before taking action” or “Hand off to you”.
2 more protections are built in. Auto-review checks actions that could affect your accounts or share information, and OpenAI says: “Certain sensitive tasks, such as changing a password, always stay with you.” The background research a dot does on its own uses read-only tools that “can't send messages, change app content, or control your browser or computer”.
None of that replaces your review. OpenAI's own caveat: “Dots can still make mistakes, so always review consequential work.”
Before you connect client files. A dot “can also review connected information proactively and form memories from it.” And unplugging an app doesn't undo that: “Disconnecting an app does not delete information your dot has already obtained from it. To delete that information, you need to delete your dot.” On training, OpenAI says content from Business, Enterprise and Edu workspaces isn't used to train its models by default; on personal plans, it's a setting, so check yours first. It adds: “We don't train directly on proactive research or your dot's notes to itself.”
Our read
What dots add is the standing job, not the abilities. TechCrunch noted that much of this “was already possible through Codex”, and that dots bundle it into 1 package. That's the point for your team: the value is in the jobs you repeat every week, which an always-on assistant keeps doing without you asking again.
Start with the job that needs the least access. That's job 1. It only needs your transcripts, and every post waits for your OK. Add your email and calendar once you trust what it drafts.
Decide where it stops before you hand it a job. 1 commenter under Futurepedia's video put it best: “you need to know what it can do alone and where it will stop for approval.” Custom Rules is where you write that down.
Expect rough edges. In Futurepedia's test, the cloud computer was down after the host reset his dot, and a scheduled task wouldn't open on the beta app he was using. Another commenter under his video reported “lots of bugs”. Review the first week of drafts closely before you widen what a dot can touch.
Start here: 3 steps
- Set the rule first. In Custom Rules, set “Ask before taking action” for anything that sends, posts or spends.
- Connect 1 folder. Give the assistant your transcripts folder and nothing else.
- Give it job 1. Paste the prompt from job 1, and review its first drafts before you add a second job.
Sources
- OpenAI, “Introducing dots”, September 29, 2026
- OpenAI Help Center, “Getting started with your dot”, read October 9, 2026
- TechCrunch, “OpenAI launches Dots, its bubbly agentic avatar”, September 29, 2026
- Skill Leap AI, “The Biggest ChatGPT Update Yet: Meet dots”, YouTube, September 29, 2026
- Futurepedia, “I Tested OpenAI's New Personal Assistant Agent: DOTS”, YouTube, September 29, 2026, and its comments, read October 9, 2026
- Our explainer, “3 Jobs to Give ChatGPT's New Dots, and the 1 to Start With”, YouTube
Quotations are exact. Lines marked Our read are MichaelBennett.co's reasoning, not OpenAI's or the creators'.
Working out where AI actually fits in your marketing?
I write these while building the systems behind them: measurement, creative pipelines, and agents that do real work. Connect on LinkedIn and tell me what you are working on. That is where these conversations start.
Connect on LinkedIn