Artificial intelligence is moving beyond the traditional chatbot model. Instead of waiting for users to type a question and request the next step, a new generation of AI systems is being designed to work continuously toward a goal. OpenAI's latest example is Dots, a new class of always-on AI agents introduced at the company's Developer Day on September 29, 2026.
OpenAI describes Dots as agents designed to work on a user's behalf, maintain context, use connected applications, and continue working toward objectives with limited supervision. The company says Dots are powered by GPT-6 Astra, operate through their own cloud computers, and can connect to more than 4,000 apps through OpenAI's plugin ecosystem.
The announcement is significant because it represents a shift from AI that primarily answers questions to AI that can take responsibility for multi-step tasks.
What Are OpenAI Dots?
OpenAI Dots are persistent AI agents designed to pursue user-defined goals rather than simply respond to individual prompts.
A conventional chatbot generally follows a pattern:
User asks → AI responds → user asks again → AI responds again.
An agent is intended to work differently:
User defines a goal → AI plans actions → AI uses tools → AI continues working → AI returns progress, results, or decisions requiring approval.
According to OpenAI, Dots can operate on their own cloud computers and browsers and can use applications that the user has connected. Users can also inspect the agent's computer and, with permission, allow a Dot to interact with a user's laptop.
This means an AI system could potentially continue working even when the user is not actively sitting in front of the computer.
How Do OpenAI Dots Work?
The basic concept behind Dots combines several AI capabilities into one persistent system.
First, the user gives the Dot a goal. The goal could involve research, software development, document preparation, business operations, customer feedback, or other recurring work.
The Dot then uses its underlying model, tools, browser access and connected applications to break the objective into tasks and execute them.
OpenAI says Dots are built on GPT-6 Astra and are designed to learn from feedback over time. The company also says they can work through interfaces such as ChatGPT, Slack and Microsoft Teams.
For example, imagine a software development team that receives hundreds of customer requests.
Instead of asking an AI assistant to analyze every request manually, a Dot could monitor incoming feedback, identify recurring issues, investigate smaller bugs, prepare code changes, run tests and return completed work for human review.
OpenAI has specifically described a developer scenario in which a Dot monitors customer feedback, scopes smaller improvements and bug fixes, builds and tests those changes, and prepares pull requests for review.
Why "Always-On" Matters
The phrase "always-on" is important because it changes how AI can be used.
Most AI tools are activated when a human asks for something. Persistent agents are designed around the opposite model: the user establishes an objective and the AI continues making progress.
This could be particularly useful for tasks that evolve over time.
For example, a sales proposal may need to be updated when customer requirements change. A research project may receive new data every day. A support team may constantly receive new feedback. A software project may generate new bugs and feature requests throughout the week.
Rather than starting from zero every time, an agent can maintain the context of the project and continue from where it previously stopped.
OpenAI says Dots can manage evolving projects and communicate through workplace platforms including Slack and Teams.
What Can OpenAI Dots Potentially Do?
The exact capabilities available to each user will depend on access, permissions and connected tools, but OpenAI's examples demonstrate several major categories.
Software Development
A Dot can be used as a persistent development assistant.
Instead of only generating code snippets, it can potentially monitor feedback, investigate bugs, make changes, test software and prepare work for a developer to review.
That could change the role of AI in software engineering from an interactive coding assistant into a more autonomous development collaborator.
Research and Analysis
OpenAI has also described a scientist using a Dot to monitor incoming data, rerun analyses, investigate unexpected results and update figures and explanations.
This is especially interesting for research workflows because scientific work is often iterative. New evidence can change earlier conclusions, creating a need for repeated analysis.
An agent capable of monitoring changes could help reduce repetitive work while keeping humans responsible for interpreting important results.
Business Operations
Enterprise teams could use persistent agents for administrative work such as preparing documents, analyzing information, updating proposals and coordinating workflows.
Reuters reported that OpenAI's Dots can use tools including Codex and ChatGPT Work for research, data analysis, document preparation and software development.
Personal Productivity
The concept can also be applied to everyday tasks.
An always-on agent could potentially organize information, track recurring work, prepare drafts, monitor selected services and remind users when human approval is needed.
OpenAI has presented examples involving practical tasks, including an early tester whose Dot noticed an unpaid invoice, prepared it and sent it after receiving approval.
Dots Versus Traditional Chatbots
The biggest difference is not necessarily intelligence alone. It is autonomy.
A chatbot generally waits for instructions.
An agent can be configured to pursue an objective through multiple steps.
Consider the difference between these two requests:
Chatbot:
"Write a sales proposal based on this information."
Agent:
"Track this customer opportunity, monitor changes in requirements, update the proposal, identify missing information, and prepare the revised version for my approval."
The second workflow requires ongoing context, tool use, decision-making and the ability to act over time.
That is the core idea behind agentic AI.
OpenAI Dots and Connected Apps
One of the most important parts of Dots is their ability to operate across applications.
OpenAI says Dots can connect to more than 4,000 apps through its plugin ecosystem. The company also says they can be reached through ChatGPT, Slack and Teams.
This matters because most real-world work does not happen inside a single application.
A marketing workflow may involve email, spreadsheets, analytics, documents, project management and communication tools. A software team may use source control, issue trackers, CI systems, messaging platforms and cloud infrastructure.
An agent that can move between those systems could potentially automate much more complex processes than an AI model operating inside a single chat window.
Human Approval Still Matters
Autonomous AI creates an important question: How much control should the user give the agent?
OpenAI says users can configure rules governing what Dots may do and when they must request permission. The company specifically says sensitive actions, including password changes and permanent data deletion, require explicit user consent.
This approval model is important because AI agents can interact with real systems.
Generating a paragraph is relatively low-risk.
Changing an account password, deleting information, sending money, modifying a production system or sending an important business communication is very different.
The more authority an AI agent receives, the more important permissions, auditability and human oversight become.
Privacy and Security Questions
The move toward always-on AI also creates new privacy and security challenges.
If an agent is expected to understand a user's goals, it may need access to business information, communications, documents and other applications.
OpenAI says Dots operate using their own cloud computers and has highlighted safeguards and permission controls. Reuters reported that OpenAI reiterated that business customer data is not used to train its AI by default, while personal-plan users can control whether conversations and Dot work are shared for training.
However, the broader AI-agent industry is facing increasing scrutiny.
On September 30, 2026, Reuters reported that the U.S. Federal Trade Commission had begun an industry-wide investigation involving Anthropic, OpenAI and other AI organizations to examine potential consumer risks associated with AI technologies and autonomous agents.
That means the development of autonomous agents is happening alongside an expanding debate about security, accountability and consumer protection.
Why OpenAI Is Entering the Agent Race
OpenAI is not developing Dots in isolation.
AI companies are increasingly competing to build systems that can perform tasks instead of simply generating text or images. Reuters described OpenAI's Dots launch as part of a broader industry race toward autonomous AI, including competition with Meta's Muse.
The strategic significance is straightforward: an AI assistant that becomes part of a user's everyday workflow could be more deeply integrated into work than a chatbot that is used occasionally.
For businesses, the potential value comes from automation.
For individuals, the attraction is time savings.
For AI companies, persistent agents create a new platform around which applications, tools and services can be connected.
What Could Dots Mean for the Future of Work?
If autonomous agents become reliable enough, the workplace could gradually shift from software that people operate manually toward software that people supervise.
Instead of opening ten applications and completing ten repetitive tasks, a worker might define the desired outcome and allow a team of specialized AI agents to execute much of the process.
OpenAI says it is also previewing specialist Dots with their own identities, access-management capabilities and deeper integrations into organizational systems.
That points toward a future in which organizations could deploy different agents for different responsibilities.
One agent could focus on customer support.
Another could assist with engineering.
Another could monitor financial reporting.
Another could manage research workflows.
Humans would still define goals, establish permissions, evaluate results and make high-impact decisions.
The Main Challenges
The promise of always-on AI comes with substantial technical challenges.
Reliability
An agent that works continuously cannot simply be occasionally impressive. It needs to perform reliably over long workflows.
OpenAI's Developer Day demonstration itself experienced technical problems, with repeated failures during voice interactions. Reuters reported that the live demos encountered glitches.
Accuracy
An AI agent making a mistake in a chat response is one thing. An AI agent making a mistake while interacting with external systems can have much larger consequences.
Security
Every connected application can potentially become another security boundary.
Permissions
Users need clear controls over what an agent can read, change, send or delete.
Accountability
When an autonomous agent makes an incorrect decision, organizations need to understand what happened and determine who or what was responsible.
These challenges are likely to become increasingly important as AI moves from content generation into action.
When Can Users Access OpenAI Dots?
OpenAI says Dots are beginning to roll out across Pro, Business Premium and Enterprise plans in eligible markets, with broader availability planned.
Availability, capabilities and limits can change as the product develops, so users should check OpenAI's current product documentation and account availability before relying on a particular feature.
Are OpenAI Dots the Future of AI Assistants?
Dots illustrate an important transition in artificial intelligence.
The first major wave of generative AI made it easy to ask a machine for information, text, images, code and ideas.
The next wave is increasingly focused on machines that can execute workflows.
That distinction could be more important than simply making language models larger or faster.
An AI system that can understand a goal, maintain context, use software, react to new information and request approval when necessary begins to resemble a digital coworker rather than a traditional chatbot.
The technology is still developing, and real-world reliability, security and user trust will determine how quickly these systems become mainstream.
Final Thoughts
OpenAI Dots represent a significant step toward persistent, autonomous AI agents.
Powered by GPT-6 Astra and designed to operate through connected applications and cloud computers, Dots are intended to continue working toward goals rather than waiting for a new prompt every time. OpenAI's examples span software development, research, business operations and personal productivity.
The bigger story is not simply the launch of another AI product. It is the changing relationship between people and software.
The traditional model is:
People operate software.
The emerging model is:
People define goals, while AI agents operate software under human supervision.
Whether that becomes the dominant model for digital work will depend on reliability, safety, privacy, cost and user trust. But with OpenAI, Meta and other companies investing heavily in autonomous agents, the shift from conversational AI toward action-oriented AI is already becoming one of the defining technology trends of 2026.