OpenAI has launched ChatGPT Work, a new agent-based mode inside ChatGPT that turns the platform from a conversational assistant into a task-running workspace. Released on July 9, 2026, the feature is designed to take a goal, gather context from connected apps and files, plan the work, and return completed outputs such as documents, spreadsheets, slides, reports, and live websites.
The launch marks one of OpenAI’s biggest product shifts since the rise of ChatGPT itself. Instead of asking users to write a prompt and refine the response, ChatGPT Work is built around delegation. A user can describe the outcome they want, approve the plan, and let the agent complete the task across multiple tools.
ChatGPT Work is powered by OpenAI’s new GPT-5.6 model family, which also launched on July 9. The family includes Sol, the flagship model for complex reasoning and long agentic workflows; Terra, the everyday workhorse; and Luna, the faster, lower-cost option.
From Chatbot to Work Agent
ChatGPT Work is designed for longer, multi-step jobs. It can connect to files, apps, messages, calendars, spreadsheets, and business systems, then use that context to complete work with less back-and-forth.
That includes preparing reports, updating slide decks, analyzing customer data, building dashboards, summarizing team conversations, and creating websites that can update as the underlying data changes.
The feature also brings together ChatGPT, Work, and Codex inside a rebuilt desktop app for Mac and Windows. Codex is now part of the broader ChatGPT experience, while OpenAI is moving away from its separate browser experiment.
The larger direction is clear. AI platforms are moving beyond chat windows and into workflow execution. The next race is not only about which model gives the best answer. It is about which system can complete the most useful work.
Key Features
One of the most important features is scheduled tasks. Users can ask ChatGPT Work to run once, repeat on a schedule, or watch for a specific trigger. For example, it can review new team messages every week and update a document or presentation when something changes.
Another major feature is Sites, currently in public beta. This allows ChatGPT Work to publish dashboards, project trackers, and interactive reports to live URLs. These are not limited to static pages. The sites can include front-end and back-end logic and continue updating as connected data changes.
On desktop, ChatGPT Work can also use computer control. It can operate other apps by clicking, typing, moving files, and navigating interfaces while the user continues working elsewhere. The user can see a second cursor moving on screen as the agent performs the task.
OpenAI has also introduced a unified plugin directory with more than 1,400 connectors, covering tools such as team chat apps, cloud storage, email, calendars, CRMs, and productivity platforms. Users can mention the app they want, and ChatGPT Work decides which connector is needed.
Safety and Approval Controls
Because ChatGPT Work can interact with connected tools, OpenAI has added approval and review controls. Plan mode shows the steps before a task begins. Check-ins pause longer jobs when the agent needs user input. Action approvals are required for more sensitive moves, such as sending messages, publishing content, or making external changes.
These controls are central to the product because agentic AI becomes more powerful and riskier when it can act across real business systems. The goal is to let users delegate work without giving the agent unlimited authority.
OpenAI says early testing showed strong results in blocking attempts to extract protected data through connected tools, but real-world use will be the larger test.
Early Business Use Cases
Several early customers have used ChatGPT Work for sales, finance, operations, event planning, and product management. In sales workflows, the agent has been used to connect CRM and email data to identify missed pipeline opportunities. In finance, it has helped turn variance analysis into updated spreadsheets, presentations, and interactive reports.
Other examples include conference planning, product launch checks, competitive analysis, and turning discovery calls into proof-of-concept materials. The common theme is that ChatGPT Work is aimed at tasks that usually require multiple tools, several hours of coordination, and repeated manual updates.
Pricing and Rollout
ChatGPT Work is rolling out first to Pro, Enterprise, and Edu users, with Plus and Business access following shortly after. The rebuilt desktop app is available globally across plans, including Free, with Chat, Work, and Codex included in one experience.
Business pricing starts at a monthly per-user rate, while Pro includes higher tiers for more demanding use. A premium Ultra mode can run multiple agents in parallel on the same task. Usage is metered through a shared agent consumption pool that also covers related work features.
What It Means for Writers and SEO Teams
For writers, bloggers, and SEO professionals, ChatGPT Work changes the workflow. The old pattern was prompt, edit, and publish. The new pattern is brief, review, approve, and publish.
That creates both pressure and opportunity. Generic content production becomes easier for everyone, which means basic AI drafting is no longer a competitive advantage. At the same time, writers who know how to brief agents using real data, customer insights, analytics, and first-hand experience can produce stronger content faster.
The skill that matters most is shifting. Writing is still important, but briefing, judgment, verification, and editorial responsibility now matter more. ChatGPT Work does not remove the need for writers. It raises the bar for what useful writing work looks like.
