OpenAI’s New Dot Agent Explained: How It Works, Costs and What It Means for Investors

The most interesting thing about OpenAI’s new Dots agent is what happens after you stop talking to it. Announced on September 29, Dots can take on ongoing responsibilities and keep making progress between conversations. For users, the promise is less time spent chasing unfinished work. For investors, the bigger question is whether that convenience can generate enough paying demand to cover the cost of providing it.

Let’s break down what a Dot actually does, who can use one and how its permissions work. We will also examine the competition and the financial test behind OpenAI’s move from answering questions to carrying out work.

What is OpenAI’s Dots agent?

Dots is the product name. A Dot is an individual AI agent that can work towards a goal over time using software and connected information. An agent combines an AI model with tools that let it do something with its answers, such as preparing a document or investigating a problem in a software project.

OpenAI’s new product uses GPT-6 Astra and gives each Dot its own computer in the cloud. That means the computer runs remotely rather than relying entirely on your laptop. OpenAI says its plugin ecosystem can connect Dots to more than 4,000 apps, subject to the connections and permissions available to the user.

Source: OpenAI’s “Introducing dots” announcement.

The app count describes the potential ecosystem. It does not mean every Dot automatically gets access to thousands of services or can perform every action within them. Likewise, future teams of personal Dots should not be confused with a feature already available to every customer.

The useful way to think about the product is as a continuing responsibility. Instead of repeatedly asking for a fresh update, you can give an agent a defined piece of work and explain when it should return to you.

How does a Dot work in everyday use?

A Dot can research, analyze data, prepare documents and build software. Its own cloud work can continue when your devices are switched off. If a task needs your personal computer, that access is optional and separate: your computer must be online with the ChatGPT app open.

That distinction matters. “Works while your laptop is off” applies to work that can run in the cloud. It is not a promise that an agent can reach files stored only on an unavailable device.

For an illustrative example, imagine a research analyst preparing a company earnings brief. The analyst could assign the Dot to collect the relevant filings, organize a comparison with the previous quarter and prepare a draft explaining what changed. The analyst would then review the figures and the interpretation before publication.

The benefit would come from reducing repeated collection and coordination. Whether the finished analysis is useful would still depend on the source material, the instructions and the quality of the review.

There are three separate connections to understand. A messaging channel lets you talk to the Dot. An app connection lets it use permitted information and tools from that service. A computer connection lets it work with the available files and software on that device. Connecting one does not automatically connect the others.

You can message a Dot through ChatGPT, Slack or Microsoft Teams and use a voice call to discuss its work. Ending the call does not necessarily stop work you assigned. OpenAI’s messaging documentation says calls initiated by the Dot are planned for after launch.

Texting needs more care: the announcement describes broader texting as coming soon, while the Help Center documents a limited beta for selected US Pro users. That is not evidence of general texting availability in India.

What makes Dots different from a regular ChatGPT conversation?

The main difference is continuity. A regular conversation can help you complete a task, but a Dot is designed to retain an ongoing responsibility as information and priorities change. It can coordinate background tasks while you continue speaking to it.

This does not mean ChatGPT previously had no tools or automation. The new proposition is a persistent agent that brings those capabilities together around continuing work.

For example, a one-off request might be: “Summarize this project plan.” An ongoing responsibility might be: “Keep track of this project’s deadlines and prepare an update when a change puts the launch at risk.” The second instruction needs a definition of what to watch, which sources to use and what the agent may do when something changes.

A Dot uses relevant ChatGPT memory and its own notes about preferences, decisions and ongoing work. Those notes are not a complete transcript of everything you have ever said. Switching between messaging channels does not create a fresh Dot, but the visible conversations remain separate.

Monitoring also needs explicit setup. Adding the Dot to a Slack channel does not by itself establish a monitoring task. For recurring work, users should specify the timing, time zone, end date where relevant and destination for the result, then check that the schedule was saved.

The practical lesson is straightforward: delegate a result with boundaries. “Keep an eye on everything” is harder to evaluate than “Check this approved source each weekday and draft an update when this specific condition changes.”

Who can use Dots and how much does access cost?

Dots is rolling out gradually. An eligible subscription does not guarantee that the feature appears immediately in an account.

Sources: OpenAI’s “Meet dots” documentation, “Getting started with your dot” Help Center article and ChatGPT Business release notes.

The Business Premium prices are subscription prices in US dollars. They are not standalone Dot fees or a confirmed India checkout quote. Users should check the price and availability shown for their account rather than convert a headline dollar price and assume it is the final bill.

India is not among the regional exclusions listed for personal Pro access. That places eligible Indian users within the stated geographic scope, but actual access still depends on age requirements, plan eligibility and rollout. Workplace users also need to follow their administrator’s settings.

There is another distinction between access and capacity. Conversations with a Dot do not count towards ChatGPT usage limits, but tasks it starts or manages in ChatGPT Work or Codex count towards those products’ limits. OpenAI includes an allowance for deeper work and has announced extended limits for the first month after launch.

Therefore, an included Dot should not be described as unlimited free computing. The subscription gives access to the agent alongside specified work allowances.

Setup starts in the ChatGPT desktop app or a desktop browser. After setup, mobile-app access depends on the supporting update being available. Mobile web is not supported at launch.

How much control does a Dot have over your work?

The word “proactive” can make the product sound more autonomous than its documented permissions allow. There are two different activities to separate.

Proactive research uses read-only tools to examine connected information and identify useful next steps. That research cannot itself send messages, change connected-app content or control a browser or computer. Assigned work can involve actions, including recurring actions, but the applicable permissions and approvals still govern them.

Before an action affects accounts or shares information, automatic review checks it against instructions, permissions, custom rules and safety requirements. Asking for a draft does not authorize sending it. Some steps, such as changing a password, remain with the user.

OpenAI also describes a protected cloud workspace, secure sign-in flows and monitoring that can pause concerning activity. These are safeguards rather than a guarantee that every action will be correct. A malicious instruction hidden inside an email or webpage could try to redirect an agent, a risk known as prompt injection.

For business use, the consequence matters as much as the frequency of errors. A typo in an internal draft is relatively easy to fix. Sharing confidential information with the wrong recipient can be much more costly. That is why an employer should evaluate which responsibilities can proceed independently and which need review.

Stopping work also has separate controls. Pausing the Dot’s current main task does not automatically stop every delegated task or cancel future schedules. Those need to be reviewed separately and stopping work does not undo actions already completed.

Privacy settings deserve attention too. OpenAI says it does not use content from Business, Enterprise or Edu workspaces to improve models by default. Personal-plan users have data controls. The company does not train directly on proactive research or private agent notes, but information from them can inform an eligible conversation or task under the applicable settings.

Is OpenAI’s Dots agent better than Meta’s Muse?

Meta introduced Muse on September 8, 2026. It is a personal AI agent powered by Muse Spark that runs on a dedicated cloud computer and can continue working after a user closes the app. The similarity shows why Dots should be viewed as part of a competitive agent market rather than an uncontested invention.

Sources: OpenAI’s “Introducing dots” and messaging documentation; Meta’s “Introducing Muse” announcement. 

The comparison supports a narrower conclusion than “one agent wins.” OpenAI emphasizes connections to professional tools and its existing work environment. Meta offers a familiar communication route through WhatsApp. Those approaches could attract different users without proving a difference in real-world reliability.

An investor should ask which product gets used repeatedly and which one produces work customers accept. An impressive demonstration is useful evidence of capability, but it does not establish retention, profitable delivery or success across thousands of messy customer workflows.

OpenAI has also previewed specialist enterprise Dots with dedicated identities and access for defined organizational responsibilities. These are starting with focused pilots. Its announced work with Microsoft on Agent 365 integration is a direction under development, not proof of a completed company-wide deployment.

Why is OpenAI launching an always-on agent?

Our assessment is that Dots gives OpenAI a way to compete for a larger part of a customer’s working day. A service used for occasional questions has a different economic role from one responsible for keeping a proposal, research project or operational process moving.

There are three potential commercial benefits. First, useful ongoing work could make customers more likely to retain a subscription. Second, demanding workflows could encourage upgrades or additional usage purchases. Third, an agent that learns a customer’s processes could become harder to replace because moving would involve rebuilding connections and working preferences.

These are possible outcomes, not disclosed Dot results. OpenAI has not provided Dot-specific revenue, customer retention or delivery margins in the launch materials reviewed for this article.

The cost side can grow alongside the revenue opportunity. Completing a task may require repeated model calls, browser steps, software execution and checking. A customer who uses an agent more often may become more valuable, but also more expensive to serve.

That makes the business test different from simply counting messages. The useful measure is how much accepted work the system produces relative to the cost of producing it. More activity only creates value when customers want the result and the company can deliver it economically.

Can a Dot justify its cost? A simple productivity model

The fairest test for a customer is time saved after review and corrections. Counting every minute of an automated draft as a productivity gain exaggerates the benefit if someone must spend almost as long checking it.

Consider an illustrative monthly-billed Business Premium subscription at the published $125 price. Assume a worker’s usable time is worth $25 an hour and the agent-assisted tasks would otherwise require 10 hours a month. Neither the hourly value nor the hours saved are measured Dot results.

The calculation is: value of net time saved equals time potentially avoided minus review time, multiplied by the assumed hourly value. Under these assumptions, the full subscription breaks even at five net hours saved each month: $125 divided by $25.

This model deliberately charges the full subscription against the benefit being tested. An existing subscriber may face no separate charge for the first Dot, while a customer upgrading from another plan should compare the additional cost. Other subscription benefits could also matter.

Saved hours are not automatically cash savings. A salaried employee may use the time to serve more customers or improve output while payroll stays the same. The model therefore estimates the value of usable capacity, not a guaranteed reduction in expenses.

For a small team, that capacity can still be meaningful. But repeated corrections can quickly consume the benefit. The strongest early use case is likely to be work that repeats, has accessible source material and produces an output that is easy to check.

What could Dots mean for OpenAI’s valuation?

OpenAI’s company-wide financial context is substantial, but it needs careful labeling. On March 31, 2026, the company announced a closed funding round with $122 billion in committed capital at a post-money valuation of $852 billion.

On September 29, Reuters reported that OpenAI was seeking at least $30 billion in fresh funding at approximately $1.4 trillion before the new money. Reuters attributed those fundraising details to Bloomberg and said discussions were at an early stage with terms subject to change.

Sources: OpenAI’s March 31 funding announcement; Reuters’ September 29 fundraising report citing Bloomberg; Reuters’ September 29 revenue report, which followed Axios’ initial reporting.

The distinction is essential: $1.4 trillion is a reported valuation being sought, not a confirmed market capitalization. The revenue run rate is also a company-wide figure. Dots had only just launched and cannot explain revenue accumulated before its introduction.

Using $70 billion as a rounded analytical denominator, a $1.4 trillion valuation implies roughly 20 times annualized revenue. This is an approximate valuation-to-run-rate comparison, not a conventional multiple based on audited full-year sales. It says nothing by itself about profit.

An annualized run rate estimates what a recent revenue pace would produce over a year if maintained. It is different from adding up the revenue actually earned during that year. A fast-growing business can have a high current run rate while its recorded annual revenue is much lower.

To see why profitable conversion matters, consider two hypothetical margins on that rounded revenue base.

Sources: Reuters’ September 29 fundraising and revenue reports. 

These are illustrations, not OpenAI earnings forecasts or a current P/E ratio. They show how strongly the valuation argument depends on converting revenue into profit, rather than growing sales alone.

Dots could strengthen the case for future revenue growth if customers pay for dependable work. It could also raise delivery costs if usage expands faster than revenue. Investors need evidence of both customer value and sustainable margins before treating the product as a justification for a higher valuation.

What does Dots launch mean for US and Indian investors?

OpenAI remains a private company in the reporting reviewed here. Its proposed funding valuation should not be confused with a price at which an ordinary investor can buy an exchange-listed OpenAI share.

Listed companies can provide different kinds of exposure to the broader business. OpenAI’s March announcement identifies Microsoft, Amazon, NVIDIA and SoftBank among its strategic funding partners. Its infrastructure ecosystem also includes cloud and chip suppliers. These connections create potential economic exposure, but each company has a much larger set of businesses and commitments to assess.

For infrastructure suppliers, successful agents could increase demand for computing. That is an inference from the work agents perform, not a disclosed Dot revenue forecast. More demand can support sales while also requiring significant investment to deliver the capacity. Supplier profits depend on pricing, utilization and the returns on that investment.

For Microsoft, the specialist-Dot collaboration adds a possible enterprise management role alongside its existing OpenAI relationship. For Meta, Muse represents a competing approach. Neither connection means a particular share price must rise after the announcement.

Indian investors evaluating US stocks should therefore separate a company’s verified relationship with OpenAI from the financial size of that relationship. A new product announcement alone is not enough to calculate its contribution to earnings or a fair share price.

For Indian businesses and professionals, the first question is more practical: which recurring responsibility is expensive to coordinate and easy to review? A research team might test source collection and comparison drafts. A services firm might test internal project updates. These are proposed trials rather than evidence of deployments or measured productivity gains.

A sensible trial would record the time previously spent, the time needed to review the agent’s output and the number of results actually accepted. That provides a better basis for deciding whether to expand access than the number of tasks started.

What should users and investors watch next?

The next stage should be judged by completed work and repeat use. Launch-day features tell us what the product is designed to do. They cannot tell us how consistently it will perform when customer data is incomplete, priorities conflict or a connected service changes.

Three scenarios capture the opportunity. In a strong outcome, review time falls and customers entrust more useful work to Dots, supporting retention and spending. In a middle outcome, the product proves valuable for a narrower set of well-defined tasks. In a weak outcome, errors and supervision consume the time saved while delivery costs remain high.

Our assessment is that Dots is a meaningful step towards AI taking on continuing work. Its financial value will depend on whether it can make that work dependable and economical. Being available around the clock is useful, but the harder achievement is producing a result worth accepting without creating an equally large checking burden.

For users, measure the hours recovered after review. For investors, watch for evidence that those hours lead to paying demand, repeat usage and improving economics. That is the connection between an appealing agent and a durable business.