OpenAI Dots: Always-on AI Agents and What They Mean for Swiss SMEs

On September 29, 2026, OpenAI unveiled 'Dots' at its DevDay in San Francisco: personal AI agents that are active around the clock and work autonomously in the background. A Dot doesn't wait for your input; it proactively pursues goals and reaches out when a decision is needed. For Swiss SMEs, this is the transition from AI assistant to autonomous digital employee – with significant opportunities and serious governance questions.

What Dots can do – and how they differ from chatbots

At its core, a Dot is an always-on agent based on GPT-6 Astra. Each Dot has its own cloud computer with a browser and can independently research, analyze data, create documents, and even build software via Codex. It's accessible through ChatGPT, Slack, and Microsoft Teams; text messages and phone calls will follow. Plugins allow connecting over 4,000 apps.

• A Dot doesn't wait for prompts; it works toward goals autonomously.
• It has its own cloud computer with a browser and can act outside chat sessions.
• It can be precisely controlled via Custom Rules: what it may do, what requires approval, what is prohibited.

The Swiss twist: enterprise-first, not for Pro users

The European launch is restricted – and Switzerland, despite not being an EU member, is grouped with the EEA and UK in a waiting zone. Pro subscribers cannot use Dots within this zone for now. Business Premium customers get Dots in all supported regions; Enterprise, Edu, and Healthcare customers can activate a beta once administrators enable it. For Swiss companies, this means: Dots arrive enterprise-first.

Opportunities and limits for Swiss SMEs

The opportunity is obvious: a Dot can take over routine tasks that currently consume valuable working time – from research to data analysis to document creation. For DACH mid-market companies, this is a lever to become operational without an in-house AI team. At the same time, autonomy introduces new risks: access credentials, approvals, and the question of liability when the agent makes a wrong decision.

How to deploy Dots responsibly

The key is not to automate everything at once, but to proceed in controlled steps. First, define clear rules: what may the Dot decide autonomously, what requires your approval, what is off-limits? OpenAI provides Custom Rules for this – use them consistently. Especially for sensitive actions like password changes or software installations, the decision remains with a human.

1. Choose a clear, low-risk use case (e.g., research or document creation).
2. Define Custom Rules: allow read access, require approval for write and send actions.
3. Review data flows: which systems and data may the Dot see? Where do they reside?
4. Start with a pilot and evaluate after two weeks: what worked well, where do you need tighter boundaries?

Another consideration is vendor lock-in. Integrating Dots into core processes makes you dependent on OpenAI – and the underlying cloud infrastructure. For Swiss companies with high data-sovereignty requirements, it's worth looking at European alternatives such as the Apertus LLM from ETH Zurich, which as an open model promises more sovereignty.

Conclusion: the next step is a conversation, not an immediate rollout

OpenAI Dots mark the shift from reactive assistants to proactive, autonomous agents. For Swiss SMEs, that means recognizing opportunities without neglecting governance. The enterprise-first rollout shows that OpenAI is initially targeting controlled corporate environments – and that's exactly where your advantage lies if you establish structures early. How to evaluate Dots concretely for your company without falling into dependencies is something we're happy to discuss in a personal conversation.