On 29 September 2026 OpenAI introduced dots, always-on agents inside ChatGPT. A dot keeps working on a goal between conversations and comes back when it needs a decision. For a small business owner that sounds like the assistant they never had time to hire. This guide covers what a dot does, where it fits, what it costs and one plan detail that matters for businesses in the Netherlands and Poland.

What a dot is

OpenAI describes dots as agents with their own cloud computer that learn from feedback over time and can work toward your goals around the clock. Through plugins they connect to over 4,000 apps, and you can reach your dot in ChatGPT, Slack or Teams (OpenAI, Introducing dots).

The ChatGPT learning pages place dots next to the tools you may already know.

Chat is for answers and drafts. ChatGPT Work is for a task you delegate now or on a schedule. Codex is for building software. A dot is for an ongoing responsibility: it tracks progress, works out the next step as things change and can hand tasks to ChatGPT Work or Codex (ChatGPT Learn, Meet dots).

That last distinction is the useful one. If you have a single job, a normal chat or a scheduled task is enough. A dot earns its place when something needs watching for weeks.

What it could take off an owner's plate

OpenAI's own examples include a content creator whose dot turns an interview transcript into clips, show notes and social posts for approval, and a sales lead whose dot keeps a proposal up to date as requirements change. It also mentions an early tester whose dot noticed a forgotten invoice, prepared it and sent it after his approval (OpenAI, Introducing dots).

Translated to a small business, good candidates look like this:

  • Following up open quotes and preparing reminder drafts when a customer goes quiet.
  • Keeping a project folder current as client feedback arrives by email.
  • Preparing a monthly summary of invoices sent and not yet paid.
  • Turning a recorded customer call into notes and a draft reply.

Each one is ongoing, lives in apps you can connect and ends with you reviewing the result.

Proactive research: helpful, and worth understanding

When you aren't working with it, a dot looks for ways to help in the background. OpenAI calls this proactive research. It uses the apps you have connected, with tools restricted to read-only, so they can't send messages, change app content or control a browser or computer (OpenAI, Introducing dots).

The help centre adds an important detail: a dot can review connected information proactively and form memories from it, even when you haven't asked about it (OpenAI Help, Getting started with your dot). In practice, connecting your mailbox means the dot learns from your mailbox. Decide on that deliberately, especially if you handle customer data.

The controls you get

The control model is clear and worth reading before you start. According to OpenAI's help centre:

OpenAI also states that approving one message doesn't give the dot ongoing permission to contact people (OpenAI Help, dots privacy, security and safety). It adds that asking for a draft doesn't authorise sending it (OpenAI Help, Getting started with your dot). That is the line we would want for any agent that writes to customers.

The tradeoffs nobody should skip

Memory you can't edit line by line. You currently can't view, delete or modify individual dot memories. To remove what a dot has learned, you delete the dot. Disconnecting an app stops new access but doesn't delete what the dot already took from it (OpenAI Help, dots privacy, security and safety).

Prompt injection remains a risk. A website, email or document can contain instructions meant to trick an agent. OpenAI describes several layers of protection and says plainly that they reduce the risk without eliminating it (OpenAI Help, dots privacy, security and safety).

Some actions can't be undone. Whether a completed action can be reversed depends on the action and the app (OpenAI Help, dots privacy, security and safety). A sent email stays sent.

Training settings depend on the plan. OpenAI doesn't use content from Business, Enterprise or Edu workspaces to improve its models by default. On personal plans, your model improvement setting decides (OpenAI, Introducing dots).

What it can't do yet

Some limits from the help centre are worth knowing before you plan around a dot. At launch it can't initiate calls to you. It can use your connected personal email account, but you can't give it its own standalone email address yet. Texting your dot is a limited beta for Pro users in the US (OpenAI Help, Getting started with your dot).

For a small business, the email point matters most. A dot that works from your own mailbox writes as you. Keep sending behind approval until you are comfortable with its tone.

Dots are also not available to users under 18 (OpenAI Help, dots privacy, security and safety). If an intern or a younger family member helps in the business, plan their work without a shared dot.

Which plan, and the EU detail

This is where many European owners will stop. Dots are included in Pro for users outside the European Economic Area, Switzerland and the UK. Business Premium users get dots across all supported ChatGPT regions, and Enterprise workspaces get a beta that an admin must switch on (OpenAI Help, Getting started with your dot).

For a business in the Netherlands or Poland, that means a personal Pro subscription isn't the route. The business workspace plan is. That has a side benefit: workspace data isn't used for training by default, which is the setting you want for client work anyway.

Your first dot is included in the plan at no extra cost, with an allowance for deeper work. Conversations with your dot don't count toward ChatGPT usage limits, but tasks it starts in Codex or ChatGPT Work do (OpenAI, Introducing dots). Check OpenAI's current pricing page for the plan cost in your country before you decide; rollout is gradual and access can take several days to appear.

How dots compare with other agents

Dots aren't the only agent with its own computer. Grok Bot works the same way, with a persistent cloud computer and approval cards, and we described its recent changes in what changed in Grok Bot since launch. If your business runs on Microsoft 365, the ownership questions from who owns the agents in your tenant apply here too. OpenAI is also working with Microsoft to bring specialist dots under the governance controls of Agent 365 (OpenAI, Introducing dots).

The honest summary: the products are converging, so choose by where your work already lives and which plan you already pay for.

A careful first month

  • Check that your plan includes dots in your country before changing subscriptions.
  • Turn on multi-factor authentication for the ChatGPT account first.
  • Connect one or two apps the dot needs for its first job, and nothing else.
  • Write Custom Rules before the first task: drafts allowed, sending and paying behind approval.
  • Give it one ongoing responsibility with a clear weekly output.
  • Review Activity View every week and note each correction.

If you want help deciding which agent fits your tools and setting rules that match how your business works, that is part of our ongoing tech partnership work.