OpenAI Dots: What the new AI agents can do – and what businesses should check first

As of 29 September 2026. OpenAI introduced “Dots” at its DevDay conference. Unlike a chatbot that answers a question and waits, a Dot is meant to pursue a task over time, use information from connected apps and get in touch when a decision is needed. OpenAI describes Dots as always-on agents. For now, that is a product claim, not proof that every complex task can be completed reliably or without oversight.

What is a Dot?

A Dot is a personal AI agent from OpenAI. According to its introduction, it runs on its own cloud computer with a browser, uses GPT-6 Astra and can connect to more than 4,000 apps through OpenAI’s plugin ecosystem. People can give it tasks, answer questions and follow its progress through ChatGPT, Slack and Microsoft Teams, among other channels. At launch, users start with one personal Dot; multiple specialised Dots are a future ambition, not a generally available feature yet.

How is this different from a normal AI chat?

The difference is not just the model but the workflow: instead of producing a single answer, a Dot is meant to carry out a multi-step assignment and, when necessary, operate software. OpenAI mentions analysing customer feedback and preparing software changes. Another example is drafting a forgotten invoice that is only sent after approval. These examples illustrate the intended workflow; they are not independently measured success rates.

Where could it help businesses?

Well-defined, repeatable tasks are a promising place to start: reviewing incoming feedback, bringing together information from different systems for a decision, or preparing a draft for a person to check. For a mid-sized business, the question is not “What could the agent theoretically do?” but: Which specific task does it take on, what data does it need, and who checks the result? A small pilot with a verifiable outcome makes more sense than blanket access to every company system.

What are the limits and risks?

An agent working across apps needs appropriate permissions. That also increases the potential impact of incorrect inputs, wrong conclusions or manipulated content from connected sources. OpenAI describes read-only access for proactive background research and controls for sensitive actions; this does not mean every deployment is automatically safe or compliant with data protection law. Businesses should limit access to what is necessary, require approval for changes and external communication, and be able to trace what the agent did. Before using personal or confidential data, review the contractual basis, data flows, retention and responsibilities.

Are Dots available in Germany?

The rollout is staggered. Reports mention ChatGPT Pro and Business Premium in eligible markets, plus an administrator-enabled beta for certain organisations. Individual Pro users in the European Economic Area reportedly do not have access at launch; Business access may be treated differently. Before planning a deployment, check availability for your own plan and country rather than assuming that a US announcement means immediate availability in Germany. Prices, usage limits and regions may change quickly.

Our take

Dots make the shift from AI answers to AI that keeps working on an assignment more visible. For businesses, the product name matters less than responsible deployment: a defined process, limited permissions, clear approval points and a measurable benefit. Whether OpenAI’s Dots are the right fit in a particular organisation can only be assessed using real tasks under the conditions actually available to it.

Sources

OpenAI’s announcement could not be accessed directly during research. Product details have therefore been cross-checked against multiple reports; access and features may change.

• OpenAI: Introducing dots ↗
• The Verge: OpenAI launches Dots ↗
• TechCrunch: OpenAI launches Dots ↗
• Engadget: Dots are OpenAI’s new personal agents ↗
• The Decoder: OpenAI Dots and availability ↗