OpenAI Dots AI Agents Launch: What the New Autonomous AI Can Do

OpenAI is taking artificial intelligence beyond chatbots with the introduction of Dots, a new kind of autonomous AI agent that is always on to achieve user-defined objectives through a slew of apps.

Dots was announced by OpenAI at its annual Developer Day in San Francisco on September 29, 2026. The launch marks the company's increased attention to AI systems that can execute more than just answers, but can perform multi-step tasks through the use of connected tools and services.

It follows a trend of tech firms racing to create “AI agents” that can understand commands, utilize software tools and make decisions with minimal human input.

What Are OpenAI Dots?

Dots are created to be independent AI agents that can persist in performing actions to achieve a goal rather than just following a series of instructions.

OpenAI's Dots are based on GPT-6 Astra and support use in apps such as Slack and Microsoft Teams, says Reuters. They can also utilize tools like Codex and ChatGPT Work to carry out tasks.

According to the documentation on OpenAI, Dots are “workspace agents” that can have associated applications, access to computers, work that can be scheduled or repeated, and more, depending on the permissions and configuration of the workspace.

This represents an important difference from a conventional chatbot.

Typically, a chatbot sits and listens for a query and then responds. An agent may be allowed to have a more general objective and can then use tools to execute more steps.

How Can Dots Work Across Apps?

A unique characteristic of Dots is that they function in the workplace.

Eligible workspace administrators will have the ability to decide if their workspace members can use Dots and which features will be enabled, according to OpenAI. Additionally, administrators can manage if Dots is allowed to be connected with Slack or Microsoft Teams.

An organisation may even have an agent to help coordinate information across various workplace tools (applications and permissions granted).

The current OpenAI workspace-agent system can be used to link agents to Slack where they can reply to messages, execute jobs with linked systems and create files. Agents may be set up to answer all messages or just when they're referred to.

But access is not automatic, it is controlled. Adding an agent to an application doesn't mean that it automatically has access to all of the services and information.

Dots Can Have Computer Access

Another important aspect is computer interaction.

According to OpenAI's documentation, Dots can also access files on a user's machine and execute commands on it in eligible Enterprise workspaces. This is disabled by default and will only work when the user's computer is on and plugged in.

If the required cloud computing abilities are available, Dots can run on Cloud computers.

This might make agents much more valuable in the tasks that require a human operator to traverse from site to site, file to file or application to application.

While computers provide added security issues, access to them also poses additional concerns. OpenAI provides administrators with controls for capabilities, connected applications and permissions.

Why AI Agents Are Becoming Important

The launch of Dots is part of a larger change taking place in the AI industry.

The first AI products and tools from generative AI were mainly for creating texts, images or code. More and more newer systems are being built to be able to act.

Rather than have an AI system ask, consider simply asking yourself:

What is the best way to analyze this data?

A user may eventually request an agent to gather the necessary information, process it, produce a report and return the outcome to a team.

According to the developer documentation from OpenAI, agents are a system that can perform tasks on its own on behalf of users without going outside the boundaries.

It might affect software development, customer service, research, administration and other work involving knowledge.

Safety Is a Major Issue

When AI systems are allowed to make decisions, there are new risks that come with this.

A wrong chat bot response could just give out incorrect information. A misconfigured agent may be able to do something on behalf of a user, access information it shouldn't or make an unwanted change.

To enable organizations to manage Dots access and capabilities, OpenAI offers controls to manage that access.

If the workspace is Enterprise, then the access to Dots is disabled by default. Administrators can enable or disable users from using Dots, connect to Slack and Microsoft Teams, access local computers and set up custom rules.

In addition, OpenAI has been introducing further safeguards to its latest agent-ready models. According to the documentation, it's able to "pause" or "stop" a conversation as a safeguard in scenarios where an agent might have misunderstood instructions, in September documentation for GPT-6 Astra.

These protections are valuable because autonomous systems can have a more significant influence than are systems that produce only responses.

What Does This Mean for Businesses?

Businesses get the greatest advantage from automation.

Workers devote considerable time to information transfer between applications, report preparation, communication tracking and repetitive digital work tasks.

AI agents can potentially take a share of these workflows as employees focus on human decision-making.

Organizations will have to be judicious in determining which agents have access to and can modify.

According to OpenAI's documentation, the access of connected apps and auth settings will dictate what an agent can do. Its advice is to restrict permissions to the extent necessary for a specific task.

Companies will more likely consider AI agents as software systems with permissions rather than using the same approach as with general chat applications.

What Does It Mean for Everyday Users?

The most significant progress for individual users is the general trend towards broader AI capabilities that can execute tasks rather than just respond to questions.

The ultimate goal is that the assistant will be able to comprehend a goal, decompose it into steps, execute those steps with various tools and deliver a complete outcome.

However this availability is dependent on product, account and workspace. According to its current documentation, some parts of the Dots features are being slowly launched and might not yet be available to all accounts.

When granting access to files, applications or computer resources, users should also be aware of what permissions an agent has.

OpenAI Dots vs Traditional Chatbots

The distinction between the two can be easily explained:

Traditional chatbot:

User questions are fed into the AI, which then provides responses.

AI agent:

User provides a goal → AI executes and plans several steps with tools provided.

Part of OpenAI's strategy towards the second model is the use of Dots.

The technology is still in the development phase and can be wrong with autonomous AI systems. But the direction is clear: AI businesses are swiftly racing to build systems that can actually do something online rather than just provide answers.

What Happens Next?

OpenAI's Dots launch will only add to the autonomous AI agent competition.

Other tech giants, such as Microsoft, Google, Anthropic, and Meta are also building tools that can use tools and complete multiple tasks.

Reliability, security, permissions, cost, and usefulness will now be the main concerns with the key questions.

When AI agents can reliably perform complex workflows without losing users' control, they could represent an important middleman between humans and the software they use all the time.

Dots is another development in the shift toward action-oriented AI agents in the industry, instead of the current trend of conversational AI.

Conclusion

This is the next phase in the progression of generative AI's evolution from answering questions to completing tasks, with OpenAI's Dots launch.

The technology may eventually yield much more effective AI assistants for use in the workplace, research, software development, and daily digital activities. However, with the capability to act, comes the importance of permissions and security, and of human oversight.

To users, the most important thing they get from the next step of AI is straightforward: It's not just about getting better answers. It's more about doing work these days.