OpenAI turns ChatGPT into a work agent for complex tasks

OpenAI presents ChatGPT Work as a new agent inside ChatGPT designed to move from advice to execution. Instead of only answering questions, it can work across connected apps, files, and workflows to produce finished materials such as spreadsheets, slide decks, documents, campaign briefs, and web apps. The core idea is that a user can give ChatGPT a broader work goal, then monitor progress, answer questions, redirect the process, and approve important actions while the agent breaks the project into smaller tasks and completes them over time.

The announcement positions ChatGPT Work as a general-purpose work agent, not just a coding tool. OpenAI says Codex technology is built into the experience, allowing ChatGPT to take action across web, mobile, and desktop environments. Codex began as a coding agent, but OpenAI says more than 5 million people now use it weekly, including more than 1 million for tasks outside software development. That is used as evidence that agent-style workflows can support business operations, sales, finance, marketing, and other knowledge work.

ChatGPT Work is powered by GPT-5.6, described in the source as OpenAI’s latest frontier model and as the system behind stronger multi-step reasoning and better use of templates and reference files. In practical terms, the product is meant to handle work where context matters: analyzing month-end budget variance, turning source materials into a marketing campaign brief, preparing for a sales meeting, or carrying customer research through multiple downstream assets. A key promise is continuity: the agent can keep context across a chain of related tasks instead of treating each request as isolated.

OpenAI also emphasizes longer-running and asynchronous work. With Scheduled Tasks, ChatGPT Work can keep projects moving when the user is away from a computer or phone, such as converting new Slack or Microsoft Teams messages into updated documents or slides and then sharing important changes with a team. The examples in the announcement focus on time saved inside OpenAI: sales teams reportedly used it to turn a discovery conversation into a tailored proof of concept within 24 hours, while finance teams used it to reduce month-end close and forecasting work from days to hours by finding data, moving it into Excel or Sheets, reconciling results, making slides, and checking outputs.

Availability is staged across plans and platforms. ChatGPT Work is rolling out on web and mobile for Pro, Enterprise, and Edu users first, with Plus and Business following over the next few days. The desktop app is presented as a stronger environment because it can use local files and apps, while a built-in browser helps with web-based tools and online files. OpenAI also says the Codex app is merging into the new ChatGPT desktop app, keeping Codex as a coding agent while adding workflow features such as inline editing in diffs, pull request review in a side panel, faster computer use, and support for multiple repositories in one project.

The announcement ties the product to plugins as the way to connect workplace context. Plugins can link ChatGPT to Slack, Microsoft Teams, Google Drive, SharePoint, email, calendars, CRMs, project trackers, and other internal systems. Users can let ChatGPT infer which plugin to use from the prompt or explicitly reference an app with an “@” mention. For organizations, the strategic shift is clear: ChatGPT Work is being framed as an operating layer for everyday business workflows, where AI can gather context, create deliverables, and coordinate work across tools.