What Is OpenAI Dots? Always-On AI Agents and What You Need to Know
Most AI assistants stop working the moment the conversation ends. You ask, they answer, and the next step waits until you come back. In September 2026, OpenAI introduced dots, always-on agents inside ChatGPT that keep working toward a goal after the chat is over, using their own cloud computer, a browser, and the apps you connect.
This blog explains what OpenAI dots are, how they work, who can use them today, which safety controls come with them, the risks enterprises still need to plan for, and how to decide whether always-on agents belong on your roadmap. If you are weighing a pilot, the CodersArts team can help you scope it safely.
OpenAI Dots at a Glance
What Does Always-On Actually Mean?
A dot is an agent that keeps working on goals you delegate, between conversations. You state a goal and the scope of what you are handing over, and the dot works on its own cloud computer with its own browser and your connected apps, messaging you when it needs input. OpenAI describes dots as powered by GPT-6 Astra. Always-on does not mean the agent acts on everything automatically, because permissions and approvals still decide what it may do.
How Is a Dot Different From a Regular ChatGPT Conversation?
A normal chat is reactive: nothing happens until you type. A dot is persistent. It can handle several projects at once, run reminders and recurring tasks, do background research when you are not working with it, and be reached in ChatGPT, Slack, or Teams, including by voice. Conversations with a dot do not count toward normal ChatGPT usage limits, while tasks it starts in Codex or ChatGPT Work count toward those tools as usual.
Why Are Enterprise Teams Paying Attention?
The shift is from prompting for an output to delegating ongoing responsibility. OpenAI is also previewing specialist dots for organizations, each with its own identity, credentials, and access permissions, and says it is working with Microsoft to connect them to Agent 365 governance controls. That puts always-on agents in the territory of security, identity, and compliance teams, not only individual productivity.
The Real Cost of Treating an Always-On Agent Like a Chatbot
A chatbot gives one answer you can read before acting. An always-on agent can read mail, edit documents, and take steps while you are away. Teams that connect it to work systems without scoping, approvals, and review often discover too late what it was allowed to touch.
Thinking about an agent pilot? CodersArts can help you decide which tasks are safe to delegate first. Write to contact@codersarts.com to discuss a readiness review.
How OpenAI Dots Work
A dot is not a single feature. It combines a model, a managed computer, connected apps, and a set of permission controls, and each part affects what the agent can do for your business.
Its Own Cloud Computer and Browser
Each dot runs on a cloud computer that is separate from yours, with its own browser, and you can open that screen to see what the dot is doing. For supported websites, a dot can use saved passwords without exposing them to the model.
Connected Apps Through Plugins
Dots connect to more than 4,000 apps through plugins, and you review each app's permissions before connecting it. Disconnecting an app does not delete information the dot already collected from it. Removing that information requires resetting the dot.
Reaching a Dot Where You Already Work
You can reach a dot in ChatGPT, Slack, or Teams, including by voice call, and context carries across channels. Dots are created in the ChatGPT desktop app or a desktop browser, and supported mobile apps can message an existing dot afterward.
Proactive Research in the Background
When you are not working with it, a dot can look for useful information using read-only tools. In that mode it cannot send messages, change content in your apps, or control your browser or computer. This separation between gathering information and taking action is central to the design.
Memory and Learning From Feedback
A dot receives your ChatGPT memories, can create its own from connected apps, and learns your preferences from feedback over time. Resetting a dot deletes its conversations, saved memories, and scheduled tasks, and then you create a new one.
Optional Connection to Your Own Computer
Access to your local computer is optional and off by default. If you allow it, a dot can create Work or Codex tasks, use local skills, and use your local browser when its cloud browser is blocked. The computer must stay online with the ChatGPT desktop app open, and you can revoke access at any time.
What Businesses Can Delegate to a Dot
OpenAI's own examples, together with the patterns we see in agent projects, point to work that is ongoing, spread across several apps, and easy for a person to review.
Software Delivery Support
A dot can turn customer feedback into proposed code changes with tests, then hand them to an engineer for review before anything is merged.
Keeping Documents Aligned With Changing Scope
When requirements change, a dot can revise launch materials, proposals, and test plans so that related documents stay consistent.
Recurring Analysis and Reporting
A dot can rerun analyses when new data arrives and refresh the figures in a report, which suits weekly or monthly reporting cycles.
Finance and Back-Office Follow-Up
A dot can notice a missed invoice, prepare it, and send it only after a person approves. OpenAI says internal testing of specialist dots covered areas such as procurement, invoice processing, email marketing, customer support, and commercial contracting.
Drafting for Approval
Posts, summaries, and routine emails can be drafted in the background and queued for a person to approve, which keeps the human decision where it matters.
Availability, Plans, and Current Limits
Availability is changing quickly, so treat the details below as a snapshot of the launch period and confirm the current status with OpenAI before planning.
Who Can Use Dots
Dots are rolling out to Pro and Business Premium users, and the first dot is included at no extra cost on those plans. Enterprise customers, including Edu and Healthcare, can join a beta once a workspace administrator enables it.
Where They Are Available
At launch, Pro access was rolling out outside the European Economic Area, Switzerland, and the United Kingdom, while Business Premium was offered in supported ChatGPT regions. Regional limits matter for multinational teams and for data residency planning.
Administrator Control
For Enterprise workspaces the beta is off by default and enabled by an administrator. Texting is not available in Business or Enterprise workspaces.
What Dots Cannot Do Yet
At launch, a dot cannot start calls to you, does not have its own standalone email address, and cannot be created on mobile. OpenAI also notes that dots can make mistakes, including when following your rules, so consequential work should be reviewed.
Safety, Permissions, and Governance Features
OpenAI has built several controls into dots. Understanding them helps you decide where its defaults are enough and where your own rules are needed.
Separating Reading From Acting
Gathering information and taking external actions are handled differently, and background research uses read-only tools. Actions that change something outside the dot go through additional checks.
Auto-Review of Planned Actions
Auto-review checks actions that could affect your accounts or share information against your instructions, your Custom Rules, and safety requirements. It decides what can proceed, what needs your approval, and what you must do yourself.
Custom Rules
You can set how a dot handles an action: take action without asking, take action if you pre-approved it, ask before acting, or hand the task to you. Rules do not grant access to apps or your computer, and they cannot override built-in safeguards.
Actions That Stay With You
Some tasks always stay with the user, such as changing a password. This gives a hard boundary for the most sensitive account actions.
Activity View and Monitoring
Activity View shows progress, including background work, and lets you redirect a dot. A monitoring system can pause or stop a dot if it detects a safety concern.
Privacy and Data Use
Content from ChatGPT Business, Enterprise, and Edu workspaces is not used to improve models by default. On personal plans, you control whether a dot's conversations and work can be used for improvement. OpenAI says it does not train directly on proactive research or a dot's notes to itself.
Risks Enterprises Should Plan For
Built-in controls reduce risk, but they do not remove it. These are the issues we would raise in any enterprise review of always-on agents.
Prompt Injection Through Connected Content
An agent that reads email, documents, and web pages can encounter text written to manipulate it. A hidden instruction in a message or file may try to push the agent toward an action you never intended. Treat everything the agent reads as untrusted input, and require approval for actions that send data out or change records.
Over-Broad App Permissions
Connecting many apps with wide access makes any mistake or manipulation more costly. Connect only what a task needs, and use the narrowest permissions each app allows.
Persistent Memory and Data Retention
Because information already collected stays with the dot after an app is disconnected, and removal requires a full reset, teams should decide in advance what may be connected and how long a dot is kept.
Shadow Adoption
Individual plans make it easy for employees to create a dot and connect work accounts without review. Set a policy on which accounts may be connected, and communicate it before the tool spreads.
Compliance, Residency, and Auditability
Regional rollout limits, text messaging through a third-party provider, and the need for records of what an agent did all affect regulated teams. Activity View helps, but many organizations will want their own logs and review process.
Reliability and Compounding Mistakes
An agent that works for long periods can build on an early wrong assumption. Review consequential work, check outputs against sources, and start with tasks where a mistake is easy to catch and reverse.
Worried about prompt injection or over-broad access?
Our engineers run security reviews and adversarial tests on agent setups before they touch real systems. Email contact@codersarts.com to discuss a review.
Why OpenAI Dots Matter for Your AI Investment
The Business Case for Always-On Agents
The business case is not abstract. Work that today waits for a person to remember, check, and follow up can keep moving, which can shorten cycle times for recurring reports, document updates, and back-office follow-up.
How Costs Shape the Decision
The first dot is included with Pro and Business Premium, with allowances for deeper work and extended limits in the first month. The larger costs are often outside the subscription: time to scope tasks, configure permissions, review output, and handle work that spills into Codex or Work, which count toward their own allowances. See the cost section below for more detail.
When Always-On Agents Are Worth Exploring
They are most worthwhile where tasks are recurring, span several apps, and can be reviewed quickly. For one-off questions, a normal chat is enough. For tasks where a mistake would be costly or hard to reverse, keep a person in the loop or choose a more controlled design.
Bringing Always-On Agents Into Your Organization
Choose a Bounded First Task
Pick a recurring task with a clear owner, low risk, and an easy way to check the result, such as a weekly summary or a document refresh.
Decide Which Apps and Accounts May Be Connected
List the systems the agent may reach and the accounts it may use. Prefer read-only access at first, and keep personal and shared accounts separate.
Set Approval Rules Before Launch
Use Custom Rules to require approval for anything that sends messages, shares files, or changes records, and review the defaults rather than assuming they match your policy.
Test With Hostile Inputs
Before real use, test the setup with emails, files, and web pages that contain misleading instructions, and confirm the agent stops where it should. Red teaming and regular testing work best together here.
Assign an Owner and a Review Rhythm
Give each agent a named human owner who checks the work on a set schedule and decides when a dot should be reset or retired.
How Does Ongoing Oversight Fit Into the Process?
Once live, review Activity View, track errors and corrections, and revisit permissions whenever the task or the connected apps change. Plan for expansion only after the first task runs reliably.
Actual implementation details vary depending on your plan, the apps involved, and the maturity of your security and governance practices.
Advantages and Limitations of Always-On Agents
Where Always-On Agents Deliver the Most Value
Advantage | Details |
Continuity between conversations | Work continues after the chat ends, so follow-up does not depend on a person remembering. |
Many apps in one place | Plugins connect more than 4,000 apps, so a single agent can work across tools. |
Reach through familiar channels | A dot can be reached in ChatGPT, Slack, and Teams, with context carried across them. |
Built-in approval controls | Auto-review and Custom Rules let teams set when the agent acts and when it asks. |
Visibility into the work | Users can open the agent's computer and use Activity View to see what it is doing. |
What Are the Trade-Offs of Always-On Agents?
Limitation | Details |
Mistakes still happen | OpenAI notes that dots can make errors, so consequential work needs review. |
Prompt injection risk | Content the agent reads can contain instructions meant to manipulate it. |
Limited control over memory | Information already collected stays after an app is disconnected, and removal needs a reset. |
Rollout and plan limits | Availability varies by plan and region, and enterprise access is in beta. |
Governance gaps | Teams may need their own logging, policy, and review beyond the built-in controls. |
How Much Do OpenAI Dots Cost?
Pricing depends on your ChatGPT plan. At launch the first dot was included with Pro and Business Premium, with allowances for deeper work and extended limits for the first month. OpenAI has said that later plans will allow more dots and more capacity per dot. Beyond subscription fees, plan for the cost of setup, review time, and any Codex or Work usage the dot triggers, since those tools count toward their own limits.
Visit this page for more information: https://help.openai.com/en/articles/20001530-getting-started-with-your-dot.
Dots Compared With Other Approaches
Dots are one way to run agents, and the right choice depends on how much control, customization, and governance your organization needs.
Dots and Task Tools Such as Codex and Work
Dots can create and manage tasks in Codex and ChatGPT Work, so they act as a coordinating layer rather than a replacement. Those tasks still follow each tool's own allowances and rules.
Dots and Custom-Built Enterprise Agents
A custom agent built with frameworks such as LangGraph or the OpenAI Agents SDK gives full control over models, data handling, tools, and audit logging. A dot is faster to start and easier to maintain, but you accept the product's design and limits.
Dots and Workflow Automation Platforms
Rule-based automation is predictable and easy to audit, which suits fixed processes. Dots handle ambiguity and changing instructions, at the cost of less predictable behavior.
Building an Always-On Agent In-House
Some organizations build their own persistent agents for strict data residency or deep integration with internal systems. This requires continuing investment in security, evaluation, and monitoring.
Which Organizations Get the Most Value From Always-On Agents?
Always-on agents tend to deliver the most value for organizations that want to:
Keep recurring work moving without constant manual follow-up
Coordinate tasks that span email, documents, chat, and business apps
Pilot agents quickly before committing to a custom build
Apply approval rules to actions that send data out or change records
Prepare for department-specific agents with their own identity and access
How CodersArts Helps You Adopt Always-On AI Agents
Always-on agents are easy to try and harder to govern. CodersArts helps businesses turn early experiments into controlled pilots that security and operations teams can support.
Readiness and Use-Case Assessment
We review your workflows, systems, and policies, then recommend which tasks suit an always-on agent and which should stay with people.
Pilot Design and Permission Architecture
We define the first task, the connected apps, approval rules, and success measures, so the pilot is small, safe, and measurable.
Security Review and Adversarial Testing
We test agent setups against prompt injection, over-broad permissions, and data leakage, and we deliver findings your team can act on.
Custom Agent Development
When you need more control than a hosted product offers, we build agents on your own infrastructure using LLM development and RAG development expertise.
Governance, Training, and Ongoing Support
We help set policies, train your teams, and monitor agents after launch. You can also bring in dedicated AI and machine learning engineers to extend your team.
Ready to pilot an always-on agent the safe way?
Contact CodersArts at contact@codersarts.com to schedule a conversation.
How Does a CodersArts Always-On Agent Engagement Work?
Most engagements follow four stages. A discovery session lists candidate tasks, the systems involved, and the risks of each. A design phase defines permissions, approval rules, and measures of success. A controlled pilot runs one or two tasks with testing against hostile inputs and a clear review process. A rollout stage adds monitoring, training, and support. Each stage ends with a deliverable, such as a task map, a security review, or a working pilot, so you can decide whether to continue before committing further.
Frequently Asked Questions
Who Can Use Dots Right Now?
At launch, Pro and Business Premium users in supported regions, and Enterprise customers in a beta that an administrator must enable. Confirm current availability with OpenAI.
Can a Dot Act Without My Approval?
It can for actions your rules allow, but you decide the rules. Auto-review and Custom Rules control what proceeds, what needs approval, and what stays with you.
Does a Dot Keep Working When My Computer Is Off?
Yes for work on its own cloud computer. If you connect your local computer, that computer must stay online with the ChatGPT desktop app open.
Is My Data Used to Train OpenAI Models?
Not by default for Business, Enterprise, and Edu content. On personal plans, you control whether eligible conversations and work are used for improvement.
Are Dots Safe for Sensitive Business Data?
They include safeguards, but safety depends on your setup. Limit connected apps, require approvals, test against malicious inputs, and review the work.
What Should a Business Evaluate Before Piloting Dots?
Evaluate the task, the apps and accounts involved, data sensitivity, regional availability, approval rules, and whether your team can monitor the agent or needs a partner.
What Services Does CodersArts Offer?
Beyond always-on AI agents and other AI and RAG specific delivery and partnership work, CodersArts offers a wider range of services that agencies, businesses, and individual developers regularly rely on, whether as part of a partnership or on their own.
AI and RAG Development
Custom AI and RAG development, starting from proof of concept through to full production builds, along with broader LLM and generative AI development for businesses building AI-powered products and internal tools.
Consultation
Project consultation for businesses and agencies evaluating an AI or machine learning initiative, helping assess feasibility, recommend the right technical approach, and scope a project before committing to full development.
One-on-One Mentorship
Personalized, expert-led mentorship for developers and teams looking to build hands-on AI, machine learning, or AI engineering skills, with guidance tailored to individual or team goals and current experience level.
Dedicated Team and Team Augmentation
Dedicated AI and machine learning engineering teams, or engineers who work as an extension of an existing in-house or agency team, scaling up or down based on project needs.
Ongoing Support and Maintenance
Post-launch monitoring, optimization, and maintenance for AI and machine learning systems already in production, helping ensure performance and reliability do not degrade over time.
Job Support Services
Remote job support for developers and engineers working on live AI, machine learning, or LLM projects, including pair programming, code reviews, workflow setup, debugging, and help meeting sprint deadlines under expert guidance.
Corporate and Team Training
Structured training and workshops for teams looking to build internal AI and machine learning capability, covering hands-on implementation as well as best practices for evaluation and production readiness.
White-Label and Partnership Delivery
CodersArts also partners with agencies, consultancies, and technology companies to deliver AI and machine learning development on their behalf, whether white-label, co-branded, or embedded alongside an existing team.
Whether you are a business starting your first agent pilot, an agency looking for a delivery partner, or a developer seeking hands-on mentorship, CodersArts offers services to support your AI journey.
Reach out at contact@codersarts.com or visit www.codersarts.com to discuss your always-on AI agents or broader AI project.
Continue Exploring AI Resources
If you found this blog helpful, explore more AI, RAG, and enterprise AI resources from CodersArts AI to see how organizations are applying these systems to real world applications.
Comments