ChatGPT Work in one sentence: your new workflow super app
ChatGPT Work is an AI workflow automation agent that turns a single goal into finished work by orchestrating multi-step task automation across connected apps, online services, and local files with minimal manual intervention. This is not another chat mode; it is OpenAI’s bid to make ChatGPT the hub where work actually gets done, not merely discussed. The agent is designed to create sheets, slide decks, documents, dashboards, web apps, and other deliverables from one request, rather than forcing users to hop between tools and copy-paste results. In effect, ChatGPT Work pushes ChatGPT from being a smart notepad into a productivity operating system—and that shift has big implications for how teams design and govern their workflows.
From answers to actions: what ChatGPT Work can now do
The headline change is ChatGPT Work automation: the agent can decompose a goal into smaller tasks, stay active for hours, and carry a project from idea to completion. Users can assign an entire workflow through one request, such as analysing a monthly budget, preparing for a sales meeting, or turning research into a campaign brief. This release extends ChatGPT’s role beyond answering prompts to executing longer projects and producing team-ready materials like sheets, slide decks, documents, dashboards, and web apps from a single work request. Computer Use lets the agent click, type, and move files across supported applications and browser tools, so it can perform actions that previously required people to juggle multiple tabs and systems. For knowledge workers, this is the first credible step toward end-to-end app integration automation inside a single AI assistant.
Deep multi-step execution and persistent projects
Under the hood, GPT-5.6 powers multi-step task automation, while integrated Codex technology writes code, uses tools, and coordinates actions as projects unfold. The model is built to reason through extended workflows and create materials that follow templates and reference files. According to OpenAI, more than five million people use Codex each week, including more than one million who already use it for work outside software development. The important shift is persistence: ChatGPT Work can remain active on longer projects, ask for guidance when needed, and request approval before sensitive actions. Scheduled Tasks extend this to recurring work, letting the agent review new messages, refresh meeting documents, check dashboards for changes, and update presentations when new feedback arrives. Instead of re-prompting for every report or prep pack, teams can treat the agent as a long-running project collaborator.
Where ChatGPT Work runs: plans, platforms, and integrations
ChatGPT Work begins a phased rollout on web and mobile to Pro, Enterprise, and Edu users, with Plus and Business plans scheduled to follow over the next few days. The updated desktop app—available globally on Windows and Mac—integrates Chat, Work, and Codex in one interface across all plans, including Free, though access to individual models and usage limits still depend on subscription tier. Connections and plugins give the agent reach into Slack, Microsoft Teams, Gmail, Google Drive, SharePoint, Salesforce, calendars, CRMs, and project trackers, with “@” mentions directing it toward specific apps when needed. The desktop app also adds a built-in browser, access to local files and applications, and Computer Use capabilities so ChatGPT Work can open and refine Google Workspace or Microsoft 365 files, then move content between browser-based tools and desktop software.
Governance, Sites, and what teams should do next
Giving an AI agent this much power raises the stakes. OpenAI is targeting organisations that want ChatGPT to manage workflows rather than isolated responses, particularly work that spans several sources, tools, and stages. For organisations, ChatGPT Work is linked to ChatGPT enterprise features: admins can manage access to plugins, browser use, connected tools, network access, sensitive actions, spend controls, and usage limits, while an auto-review system checks important actions involving connected tools and APIs before they are completed. Wider access can reduce repetitive work, but it also increases the consequences of incorrect instructions, excessive permissions, or incomplete information. Meanwhile, Sites—now in public beta—turns project outputs into interactive sites or web apps that can be tested inside ChatGPT and updated when source data changes. Teams that treat ChatGPT Work as a workflow super app, not just a chatbot, will move fastest—but only if they pair automation with disciplined governance.
