TL;DR
OpenAI Dots are always-on AI agents that live inside ChatGPT and keep working after you close the chat. Each Dot runs on GPT-6 Astra and gets its own cloud computer and browser. It can reach more than 4,000 apps through plugins and talks to you in ChatGPT, Slack, or Microsoft Teams. Your first Dot comes included with ChatGPT Pro and Business Premium, and Enterprise, Edu and Healthcare workspaces get a beta that admins switch on. Custom Rules let you decide which actions it takes alone and which wait for your approval. OpenAI has not yet published what extra Dots cost, so treat it as a pilot tool for one well-defined job before you scale.
Key Takeaways OpenAI Dots are persistent ChatGPT agents that own ongoing work instead of answering one prompt. Each Dot gets its own cloud computer, browser, memory, and access to 4,000+ apps. The first Dot is included with ChatGPT Pro and Business Premium at no extra charge. Enterprise, Edu, and Healthcare get a beta that stays off until an admin enables it. Custom Rules set four levels of autonomy, from acting alone to handing the task back. Pricing for extra Dots and long-term usage allowances is still unpublished. Companies that start with one narrow, owned job will learn fastest and risk the least. Watch on YouTube
How Enterprise AI Agents Are Designed for Real Business Decisions
See how enterprise teams scope agents around real decisions, data, and approvals before giving them ongoing work.
An Invoice Nobody Asked It to Find One early tester’s Dot noticed something its owner had missed. A publication had never been invoiced for finished work. The Dot spotted the gap in his email threads, drafted the invoice, and sent it as a PDF once he approved it. In fact, nobody had assigned that task.
That small story, shared in OpenAI’s launch announcement , captures the shift. Most chatbots wait for instructions. Dots, however, are built to watch your work, notice what needs doing, and come back with something ready for your sign-off.
This guide covers what OpenAI Dots are, how they work, what they cost, who can use them today, and where they help or fall short. It also explains how enterprise teams can prepare for agentic AI that runs in the background, without losing control of their data and systems.
What Are OpenAI Dots? OpenAI Dots are named, persistent AI agents inside ChatGPT. OpenAI announced them at DevDay on September 29, 2026, and describes them as assistants that keep working across complex projects and everyday tasks.
A regular ChatGPT conversation ends when you stop typing. A Dot does not. Instead, you give it a goal and some standards, and it keeps making progress between conversations. Along the way, it checks connected apps, runs scheduled work, and returns with drafts, updates, or questions.
Why OpenAI Calls Them Always-On Agents The “always-on” label means a Dot works on your behalf around the clock, even when ChatGPT is closed. It can hold several projects at once. In addition, it can run what OpenAI calls proactive research, scanning your connected apps for useful next steps while you are busy elsewhere.
That background mode is deliberately limited. During proactive research, a Dot has read-only access. It cannot send messages, change content, or take control of your computer. As a result, real actions happen only through tasks you set up or approve.
Where Dots Fit in the ChatGPT Product Lineup OpenAI Dots sit on top of tools the company already ships. A Dot can hand deeper work to Codex for software tasks or to ChatGPT Work for longer research and documents. It can also join ChatGPT Space, OpenAI’s new shared workspace where people, ChatGPT, and Dots edit the same Pages and project files.
Think of the Dot as the persistent layer. In other words, the other tools are the hands it calls on when a job needs them. If you want the bigger picture of how these pieces connect, our guide to AI agent architecture explains the layers most agent products share.
How OpenAI Dots Work Five parts make a Dot work. The model reasons, the cloud computer acts, plugins reach your systems, channels carry the conversation, and memory keeps context from one day to the next.
1. GPT-6 Astra Does the Reasoning Every Dot runs on GPT-6 Astra, OpenAI’s flagship model. The model is what decides what to do next, how to break a goal into steps, and when a result is good enough to show you. A Dot is a product built around that model.
2. Each Dot Gets Its Own Cloud Computer A Dot works inside an isolated cloud computer with its own browser. It does not touch your laptop by default. However, you can open its screen at any time and watch the work happen.
If a task truly needs your local machine, you can grant access through the ChatGPT desktop app. That permission starts switched off. Similarly, for websites that need a login, saved credentials are used without exposing your password to the model.
3. Plugins Connect It to Your Business Apps OpenAI says its plugin library connects Dots to more than 4,000 apps. For example, that covers email, calendars, files, CRM, project tools, and developer platforms.
Here is a useful way to think about it. The model decides what to do, while plugin permissions decide what the Dot is allowed to reach. Because those two things are separate, the second one is where your IT team keeps control. The guide to context-aware AI agents with MCP covers why that split matters for enterprise systems.
4. Channels Let You Reach It From Anywhere You can talk to your Dot in ChatGPT on desktop, web, or mobile. In addition, you can message it in Slack or Microsoft Teams, and it carries the same context across all of them. However, texting is in a limited beta for US Pro users only.
Setup happens on desktop. OpenAI’s help center guide says you create your Dot in the ChatGPT desktop app or desktop browser, connect your apps, and then use the mobile app afterward.
5. Memory Builds Working Context Over Time A Dot receives your existing ChatGPT memories and creates its own, including notes from connected apps. Over time, it learns your standards, preferences, and recurring work, so you stop re-explaining the basics.
That persistence has a governance side. For instance, disconnecting an app does not erase what the Dot already learned from it. Instead, only a full reset deletes the Dot, its conversations, its memories, and its scheduled tasks together.
Key Features of OpenAI Dots Beyond the architecture, a handful of features shape how Dots feel in daily use. Of these, the ones below are what business users will notice first.
Persistent ownership. A Dot keeps several projects moving at once without you opening a new thread for each.Custom Rules. You decide, action by action, how much the Dot can do without asking.Activity View. You can inspect background work and step in at any point.Auto-review. A separate check reviews consequential actions against your instructions and safety rules before they run.Scheduled tasks. Recurring jobs run on a timetable you set.A personal identity. You can name your Dot and give it its own handle.Custom Rules in Practice Custom Rules give each type of action one of four settings. The Dot can take action without asking, act only if you pre-approved it, ask before acting, or hand the task off to you entirely.
Even so, some actions stay with you no matter what. For example, password changes always require a human. In addition, OpenAI’s monitoring can pause or stop a Dot if it detects a safety concern.
Specialist Dots for Enterprise Roles OpenAI is also previewing specialist Dots. Unlike a personal Dot, these get their own organizational identity, credentials, and system access for one defined business responsibility.
OpenAI says it has tested them internally in procurement, invoice processing, email marketing, customer support, and commercial contracting. For now, they are starting with focused enterprise pilots, and OpenAI is working with Microsoft to bring them into Agent 365 so IT can manage them with existing security controls.
What Can OpenAI Dots Do? Use Cases by Team The best way to understand a Dot is to look at the work it can own. OpenAI shared several examples at launch, and they map neatly onto common business functions.
1. Software Engineering A Dot can watch customer feedback, spot bugs, scope fixes, build and test them, and return pull requests with demo videos. For instance, inside OpenAI, a bug that appears in Slack sets a Dot investigating right away. Teams exploring this pattern can also read the Kanerika guide to AI coding agents .
2. Sales and Solutions Engineering A Dot can compare a prospect’s requirements against product documentation, list the tests a deal needs, build a proof of concept, and keep the proposal current as requirements change. As a result, sales engineers get a running head start on every evaluation.
3. Marketing and Product Launches Once a Dot learns your audience, positioning, and creative standards, it can revise launch materials when the product scope shifts. In turn, that removes the tedious step of hunting down every asset that mentions an old feature.
4. Research and Analytics For scientists and analysts, a Dot can rerun analyses as new data arrives and update the figures in a paper or report. Besides that, it can investigate unexpected results and flag what changed.
5. Content and Media A content team’s Dot can read interview transcripts, find clip-worthy moments, prepare show notes, and draft social posts in the creator’s voice. Then, when an editor makes a change, the Dot can carry it across related materials.
6. Finance, Operations, and Personal Admin The invoice story above is the personal version of this. Similarly, another demo showed a Dot spotting a calendar conflict in Slack and offering meal delivery options with prices, then letting its owner make the final pick. For a wider set of ideas, see our roundup of real AI agent examples .
Benefits of OpenAI Dots for Users and Businesses For individuals, the payoff is simple. As a result, you spend less time chasing loose ends, re-explaining context, and remembering follow-ups. Meanwhile, the Dot keeps track and comes back when it needs a decision.
For businesses, however, the gains are bigger and also harder to capture. Here is where always-on agents can pay off.
From tasks to responsibilities. You assign an outcome, such as “keep this proposal current,” instead of a one-off request.Less repeated context. Memory means standards and preferences carry forward.More work done asynchronously. Progress happens overnight and between meetings.Faster response to change. A Dot notices a new bug, a changed requirement, or a missed invoice while it is still small.Work where people already talk. Slack and Teams access means the agent shows up in existing workflows.
Even so, these benefits depend on how well the Dot is set up. For example, an agent with vague goals, broad app access, and no review habit will create noise. The Kanerika breakdown of AI agents vs AI assistants explains why ownership demands more structure than a chat assistant does.
OpenAI Dots Pricing, Plans, and Availability OpenAI Dots are not sold as a separate product yet. Instead, access comes through specific ChatGPT plans, and the first Dot is included in those plans at no extra charge.
1. ChatGPT Pro Pro now comes in three tiers at $100, $200, and $500 a month, according to OpenAI’s Pro tiers page . Currently, Dots are rolling out gradually to Pro users in eligible markets. However, Pro users in the European Economic Area, Switzerland, and the UK are excluded for now.
2. ChatGPT Business Premium Business Premium seats cost $100 per user per month billed annually, or $125 billed monthly, per OpenAI’s business pricing . In addition, they include five times the usage of a Standard seat and access to Dots in all supported ChatGPT regions. By contrast, Standard Business seats do not include Dots.
3. ChatGPT Enterprise, Edu, and Healthcare These workspaces can try Dots in beta. However, the feature is off by default, and a workspace admin has to turn it on. Meanwhile, Enterprise pricing stays custom.
How Usage Is Counted Conversations with your Dot do not count toward ChatGPT usage limits. On the other hand, tasks it starts in Codex or ChatGPT Work do count against those limits. In addition, OpenAI says it is extending limits for deeper work during the first month after launch.
Equally important is what OpenAI has not published. For example, there is no public price yet for additional Dots, for higher speed, or for more monthly work capacity. Therefore, budget owners should plan a pilot around the included Dot and revisit costs once those numbers land.
Talk to Kanerika
Planning an AI Agent Pilot?
Talk to Kanerika about where an always-on agent fits your workflows, what access it needs, and how to measure results before you scale.
Book a Meeting →
OpenAI Dots vs ChatGPT Agent vs Workspace Agents OpenAI now has several agent products, and the names overlap. This table shows how Dots differ from the two earlier ones.
How OpenAI’s agent products compare
Feature ChatGPT Agent Workspace Agents OpenAI Dots Launched July 2025 April 2026 September 2026 Work style One task per session Recurring team workflows Ongoing responsibilities, always on Built for Individual users Teams in a workspace One person’s work, with team access via Slack and Teams Memory Session context Improves through repeated use Persistent memories plus notes from connected apps Plans Paid ChatGPT plans Business, Enterprise, Edu, Teachers Pro, Business Premium, Enterprise beta
The short version is that ChatGPT agent runs a job when you ask. Workspace agents automate a shared team process. Dots take ongoing ownership of one person’s work and keep going without a new prompt. If you build agents directly on OpenAI’s platform, the Kanerika overview of OpenAI AgentKit covers the developer side.
Kanerika Service
Agentic AI Development Services
Kanerika designs and builds AI agents tailored to your workflows, data, and approval rules, connected to the systems your teams already use.
Explore Agentic AI Services
Security, Safety, and Admin Controls An agent that works while you are away needs firm limits. For that reason, OpenAI built several layers into Dots, and enterprises should understand each one before rollout.
Isolation. Each Dot runs in its own cloud computer, separate from your devices unless you connect them.App controls. Organizations choose which apps Dots can use through existing ChatGPT app controls.Approval rules. Custom Rules set what runs alone, what needs approval, and what gets handed back.Read-only background work. Proactive research cannot send, edit, or take control of anything.Auto-review and monitoring. Consequential actions get checked, and suspicious work can be paused.Sensitive tasks stay human. Password changes and similar account actions always need you.
On data use, OpenAI says workspace content from ChatGPT Business, Enterprise, and Edu is not used to train its models by default. Meanwhile, personal-plan users can choose whether Dot conversations help improve models. Furthermore, OpenAI says proactive research and a Dot’s own notes are not used for direct training.
Overall, these controls are a solid start. However, they do not replace your own policies. A strong agentic AI governance model still decides who may run a Dot, which data it may touch, and how its work gets reviewed.
Limitations and Open Questions to Watch OpenAI Dots launched only days ago. Early demos are promising. Still, they do not prove how an agent performs over weeks of real work. So here is what buyers should watch before committing.
Unpublished pricing. Costs for additional Dots, more capacity, and specialist Dots have not been announced.Regional gaps. Pro users in the EEA, UK, and Switzerland cannot get Dots yet.One Dot per user for now. OpenAI plans to let users run teams of Dots later.Hosted runtime only. Dots run on OpenAI-managed computers, which may not suit every data residency or compliance need.Lasting memory. Disconnecting an app does not erase what the Dot learned from it.Agent safety is still maturing. Days before launch, OpenAI disclosed that some of its AI agents had posted 53 user-provided images to image-hosting sites. That incident was not tied to Dots, but it shows why approval rules matter.
OpenAI’s own guidance says Dots can still make mistakes, so people should review consequential work. In fact, that is good advice for any autonomous system. The Kanerika article on common AI agent challenges explains the failure modes to plan for.
How to Prepare Your Enterprise for an OpenAI Dots Rollout The companies that get value from OpenAI Dots fastest will treat them like a new team member with a narrow job description. In practice, a careful rollout looks something like this.
Pick one narrow job. Choose a recurring responsibility with a clear human owner, such as weekly proposal updates or invoice follow-ups.Map the apps that job needs. Connect only those systems, with the least access that still works.Set Custom Rules conservatively. Default to “ask before taking action” for anything that leaves the company.Pilot and review weekly. Read the Activity View, note mistakes, and refine the Dot’s instructions.Measure before you scale. Track time saved, error rates, and turnaround, then widen access once results hold.Measurement is the step most teams skip. These guides to AI agent evaluation and AI agent observability show how to track agent quality once a pilot is live.
When a General Dot Is Not Enough Dots are general assistants built for broad work. However, many enterprise problems need something more specific, such as an agent that follows a regulated workflow, runs on private data, or connects to a legacy system no plugin covers.
OpenAI Dots vs a custom enterprise AI agent
Factor OpenAI Dots Custom Enterprise Agent Time to start Minutes, inside ChatGPT Weeks of design and build Best for Personal and team knowledge work Core business processes with strict rules Data and hosting OpenAI-managed cloud computer Your cloud, your data boundaries System access Apps available as plugins Any system, including legacy and internal tools Model choice GPT-6 Astra Chosen per use case, including OpenAI and Claude models Governance ChatGPT controls and Custom Rules Designed around your audit and compliance needs
Most enterprises will end up using both. In practice, Dots handle everyday knowledge work, while purpose-built agents run the processes that carry real business risk.
AI Assessment
Is Your Organization Ready for AI Agents?
Take Kanerika’s AI Maturity Assessment to see where your data, processes, and governance stand before you roll out always-on agents.
Start Your AI Assessment →
How Kanerika Builds Custom AI Agents for Enterprise Challenges Off-the-shelf agents like Dots raise the bar for what people expect from AI at work. At the same time, they expose a gap. After all, every enterprise has workflows, data, and compliance rules that a general assistant was never designed to handle.
That gap is where Kanerika works. The team designs and delivers custom AI solutions and agents tailored to each client’s processes, data, and controls. As part of the partner programs of AI leaders like OpenAI and Anthropic, Kanerika builds on both OpenAI’s models and Claude, so each solution uses the model that fits the job instead of locking a client into one vendor. Read more about these relationships on the OpenAI partnership and Anthropic partnership pages.
To make that work, every engagement follows a clear path.
Assess. Find the workflows where an agent will pay off, and check that the data behind them is ready.Design. Define the agent’s job, its system access, and the points where a human must approve.Build. Connect the agent to real systems such as Salesforce, Zendesk, Snowflake, or internal databases.Govern. Add audit trails, access controls, and evaluation so the agent stays reliable.Scale. Expand to new teams once the first agent proves its value.
Two recent projects show what this looks like in practice. Both projects were delivered for a global expert network that connects decision-makers with more than one million subject-matter experts.
Case Study 1: An AI Compliance Agent for Expert Vetting The client’s compliance team screened every expert manually for negative news across many sources. As a result, backlogs grew, clearances slowed, and important findings could slip through.
The Kanerika team built an AI compliance agent that connects to internal expert databases, searches news, social media, and professional records, and produces structured reports with citations mapped to compliance criteria. Under the hood, it runs on Snowflake, Python, React, and Salesforce. The results included 3x faster expert vetting, a 70% drop in backlog cases, a 40% reduction in event delays, and 60% less time spent on negative-news screening. Read the full AI compliance agent case study .
Case Study 2: An AI Member Support Agent The same organization’s executives spent hours answering repetitive member questions by searching siloed knowledge bases and old tickets. Consequently, responses were slow and member satisfaction suffered.
The Kanerika team deployed an AI support agent connected to the knowledge base and Zendesk. Today, it resolves routine questions instantly, writes ticket summaries with suggested next steps, and routes complex cases to a person. The impact was clear. Sixty-five percent of member queries are now resolved through self-service, ticket volume fell 42%, cost per ticket dropped 31%, and member satisfaction rose 25%. See the AI member support agent case study for details, and our guide to AI agents for customer support for the wider playbook.
Both agents share the traits that matter in enterprise settings. Specifically, each has a narrow job, clear system access, and a human in the loop for the decisions that carry risk. Likewise, that same discipline makes tools like Dots safer to adopt.
If you are weighing Dots, a custom agent, or both, Kanerika’s agentic AI services and AI governance services can help you choose the right mix and roll it out with confidence.
Frequently Asked Questions
What are OpenAI Dots? OpenAI Dots are always-on AI agents inside ChatGPT, announced at OpenAI DevDay on September 29, 2026. Each Dot runs on GPT-6 Astra, gets its own cloud computer and browser, and keeps working on your goals between conversations. It connects to more than 4,000 apps through plugins and checks in with drafts, updates, or approval requests in ChatGPT, Slack, or Microsoft Teams.
How do OpenAI Dots work? A Dot combines five parts. GPT-6 Astra handles reasoning, an isolated cloud computer and browser carry out tasks, plugins connect it to business apps, channels let you reach it in ChatGPT, Slack, or Teams, and memory keeps your standards and preferences. Custom Rules decide which actions it takes alone and which wait for your approval, while background research stays read-only.
Are OpenAI Dots the same as ChatGPT agent? No. ChatGPT agent, launched in July 2025, completes one task per session when you ask. Dots take ongoing ownership of work, keep memories, run scheduled tasks, and continue between conversations. Workspace agents, launched in April 2026, automate shared team workflows. Dots focus on one person’s responsibilities, with access through Slack and Teams so colleagues can work alongside them.
How much do OpenAI Dots cost? The first Dot is included at no extra charge with ChatGPT Pro, which comes in $100, $200, and $500 monthly tiers, and with Business Premium seats at $100 per user per month billed annually. Enterprise, Edu, and Healthcare workspaces get a beta. OpenAI has not yet published pricing for additional Dots, faster speed, or more monthly work capacity.
Are OpenAI Dots available on ChatGPT Plus? Not at launch. Dots are rolling out to ChatGPT Pro and Business Premium users in eligible markets, with a beta for Enterprise, Edu, and Healthcare workspaces when an admin enables it. Plus and Business Standard seats do not include Dots. Pro users in the European Economic Area, Switzerland, and the UK are also excluded for now, while Business Premium covers all supported regions.
Do OpenAI Dots keep working when ChatGPT is closed? Yes. Dots run on their own cloud computers, so they keep making progress after you close the app. They can run scheduled tasks and perform proactive research on connected apps. That background research is read-only, so a Dot cannot send messages, change content, or control your computer unless you set up or approve a task that allows it.
What apps can OpenAI Dots connect to? OpenAI says its plugin library lets Dots connect to more than 4,000 apps, including email, calendars, file storage, collaboration tools, CRM, and developer platforms. You can message a Dot in Slack and Microsoft Teams, and it can start Codex and ChatGPT Work tasks. Organizations control which apps a Dot may use through existing ChatGPT app controls.
Are OpenAI Dots safe for enterprise data? Dots include several safeguards. Each runs in an isolated cloud computer, background research is read-only, Custom Rules set approval levels, and auto-review checks consequential actions. OpenAI says Business, Enterprise, and Edu workspace content is not used for training by default. Enterprises should still apply least-privilege access, review agent activity, and set their own governance policies.
What are specialist Dots? Specialist Dots are an enterprise preview of Dots built for one defined business role. Each gets its own organizational identity, credentials, and system access. OpenAI tested them internally in procurement, invoice processing, email marketing, customer support, and commercial contracting. They start with focused pilots, and OpenAI is integrating them with Microsoft Agent 365 for governance through existing security controls.
Should enterprises use Dots or build custom AI agents? Many will use both. Dots suit everyday knowledge work inside ChatGPT and can start in minutes. Custom AI agents fit core processes that need strict rules, private data boundaries, legacy system access, or a choice of models. A practical approach is to pilot Dots on narrow personal tasks while building purpose-designed agents for regulated, high-risk, or deeply integrated workflows.
