Published on August 12, 2026
ChatGPT Work is designed to move beyond individual prompts by operating across files and applications, working on projects for hours and turning broad objectives into finished documents, analyses and other deliverables.
OpenAI introduced ChatGPT Work on July 9, 2026, as a new agent for longer and more involved projects. Instead of focusing mainly on answering a question or producing a quick draft, it can gather context, divide an objective into smaller steps, use connected tools and continue working until it produces a finished result.
According to OpenAI’s announcement of ChatGPT Work, the agent can work across applications and files, remain active on a complex project for hours and create materials such as documents, spreadsheets, presentations and web applications. Users can follow its progress, provide additional information, change its direction and approve important actions while the project is underway.
The central idea is delegation. A user provides the desired outcome, relevant materials and constraints; ChatGPT Work then coordinates more of the process required to reach that outcome. This positions it as an execution layer for digital work rather than simply another conversational AI interface.
OpenAI now distinguishes between three experiences inside its broader ChatGPT ecosystem.
Chat remains the conversational option for questions, brainstorming, searches and relatively quick assistance. Codex is dedicated primarily to software development, including writing code, working with repositories and running technical commands. ChatGPT Work sits between and across those experiences: it is intended for research, analysis and multi-stage projects that end in a completed deliverable.
This distinction matters because many professional tasks cannot be completed through a single response. Preparing a market analysis, for example, may require collecting information from multiple sources, comparing findings, organizing evidence, calculating results and building a presentation for decision-makers.
The official ChatGPT Work product page describes the agent as a system that gathers context, plans an approach and takes action across tools, files and desktop applications. It is powered by the GPT-5.6 model family, which allows it to navigate unclear requirements, adapt as work develops and produce outputs that follow supplied reference files or templates.
ChatGPT Work therefore changes the basic unit of interaction. The user is no longer limited to requesting an answer. The request can instead describe an outcome: analyze these materials, identify the important findings, build a forecast and present the conclusions in an executive-ready format.
A defining capability of ChatGPT Work is its ability to carry context through several stages of the same project.
OpenAI gives examples such as analyzing monthly budget variances, converting source materials into a marketing campaign brief and preparing for a sales meeting. A larger workflow could begin with customer research, continue with the creation of a campaign strategy and finish with marketing materials adapted for different markets.
The agent can create documents, spreadsheets, presentations, reports and ChatGPT Sites. It can also continue refining those outputs instead of treating the first draft as the end of the task. This makes it potentially useful for work that normally requires people to transfer information repeatedly between research tools, office software and communication platforms.
Connected plugins expand the context available to the agent. OpenAI says these connections can include Slack, Microsoft Teams, Google Drive, SharePoint, email services, calendars, customer relationship management systems and project trackers. ChatGPT Work can retrieve relevant information from connected sources and use it to build or update the requested output.
On the desktop application, the agent can also work with approved local folders, files and applications. Its built-in browser can gather information from the web and interact with supported online tools. Computer Use allows it to perform actions such as clicking, typing and moving files, provided the necessary permissions have been granted.
There is an important platform distinction. Work running through the web or mobile application operates in the cloud and cannot directly access files stored on the user’s computer. Local file and desktop-app access requires the macOS or Windows application and depends on the user’s plan, workspace settings and permissions. Local conversations and outputs remain on the device unless the user deliberately moves or shares them.
ChatGPT Work is designed to remain active on projects that take considerably longer than an ordinary conversation. OpenAI says it can work for hours when necessary, breaking a complex objective into smaller components and completing them independently.
That does not mean the user disappears from the process. Progress remains visible, and users can answer questions, supply missing context, request changes or redirect the agent. Important actions may require approval before they are completed.
Scheduled Tasks extend this model beyond a single work session. A task can run once at a future time, repeat on a schedule, respond to a trigger or monitor an information source for changes.
For example, ChatGPT Work could review new team messages each week and refresh a meeting agenda, check a collection of websites each morning and summarize important changes, or update a presentation when new feedback arrives by email. This allows the agent to support recurring digital workflows rather than waiting for the user to restart the process manually every time.
The broader productivity implication is a shift from real-time prompting toward managed delegation. A person can initiate a project, leave the application and return later to review the work, provided the task does not require an unresolved decision or approval.
ChatGPT Work is likely to be most valuable when a project involves several information sources, multiple applications and a clear final deliverable.
A marketing team could combine customer research, campaign performance and internal planning documents into an updated strategy presentation. A finance team could analyze operating results and convert the findings into a forecast or executive dashboard. Managers could consolidate updates from emails, project trackers and team conversations into reports containing blockers, responsibilities and next actions.
Creators and entrepreneurs could use the agent to organize research, prepare business materials or maintain recurring operational documents. Students may find it useful for longer research and presentation projects, although they still need to verify evidence, respect academic rules and ensure that the final work reflects their own understanding.
The greatest potential gain is not necessarily faster writing. It is the reduction of coordination work between stages: locating files, transferring information, reformatting outputs and repeatedly explaining the same project context to different tools.
However, not every task needs an autonomous agent. A straightforward question, short rewrite or quick calculation will generally remain better suited to the regular Chat experience. ChatGPT Work becomes more relevant as the number of steps, sources and deliverables increases.
An agent that can interact with applications and take actions also introduces more significant risks than a chatbot that only produces text.
Connected applications may contain confidential messages, customer records, financial data or internal documents. Computer Use may allow the agent to operate within systems where an incorrect action could affect files, communications or business processes.
OpenAI says users control what ChatGPT Work can access, when it should ask for approval and which actions it is permitted to perform. Business, Enterprise and Edu administrators can also manage access to plugins, company information, browser capabilities and connected systems. An auto-review mechanism is designed to examine important actions involving tools and APIs before they occur.
These controls reduce risk but do not make the agent infallible. Users should still review factual conclusions, calculations and consequential actions. Clear instructions, limited permissions and defined approval points become increasingly important as the agent receives more access and autonomy.
Usage is another practical consideration. OpenAI states that Work follows the same general usage structure as Codex, and consumption varies according to the complexity and duration of the task. Longer projects may therefore use more of a plan’s included capacity than ordinary ChatGPT conversations.
As of August 2026, ChatGPT Work is available through OpenAI’s new desktop application for macOS and Windows. OpenAI says desktop access is available across ChatGPT plans, although specific capabilities, usage limits and local access may vary.
On the web and mobile applications, Work is available to eligible paid users, including Plus, Pro, Business, Enterprise and Edu accounts. Workspace administrators may restrict or disable access, connected applications and individual capabilities.
The desktop experience is the most complete because it can combine cloud tools with local files, desktop applications, browser activity and Computer Use. Cloud Work conversations can sync between desktop, web and mobile, while local conversations remain on the computer.
ChatGPT Work represents a significant change in OpenAI’s approach to productivity. Its purpose is not simply to make individual answers more capable, but to let ChatGPT remain involved across the full duration of a project.
By combining long-running reasoning, connected applications, file access, scheduled tasks and finished deliverables, it can turn a broad objective into a coordinated sequence of work. That could reduce the manual effort involved in moving information between research, analysis, production and reporting.
Its value will depend on how reliably it performs, how carefully permissions are managed and whether the time saved is greater than the effort required to review its work. Human judgment remains essential, particularly for sensitive data, consequential decisions and external communications.
The larger implication for the future of work is that AI productivity tools are beginning to compete not only on the quality of their answers, but on how much of a project they can responsibly carry from initial goal to completed result.
OpenAI — ChatGPT Is Now a Partner for Your Most Ambitious Work — Published July 9, 2026.
OpenAI — ChatGPT Work for Every Team — Accessed August 5, 2026.
OpenAI Help Center — ChatGPT Work and Codex — Updated August 2026; accessed August 5, 2026.
OpenAI Help Center — ChatGPT Release Notes — ChatGPT Work announcement published July 9, 2026; accessed August 5, 2026.
