OpenAI launches ChatGPT Work, a new agent for long-running business tasks

news

OpenAI is turning ChatGPT into a more explicit work agent with ChatGPT Work, GPT-5.6, plugin integrations, scheduled tasks, and a merged desktop experience that now folds Codex into the main app.

OpenAI's official ChatGPT Work launch image showing desktop and mobile experiences with Google Calendar, Slack, and Gmail integrations

OpenAI launches ChatGPT Work, a new agent for long-running business tasks

OpenAI has launched ChatGPT Work, a new mode inside ChatGPT designed for longer, multi-step tasks that go beyond normal chat. The product positions ChatGPT less like a smart reply box and more like a working agent that can gather information across apps, create finished outputs, and stay with a project for hours instead of minutes.

What the official source confirms

OpenAI's official launch post says ChatGPT Work can create finished materials such as docs, slides, sheets, reports, and web apps while using connected apps and workflows as context. The company also says the product is powered by GPT-5.6, its latest frontier model for multi-step reasoning and structured output generation.

The rollout is not purely web-based. OpenAI says ChatGPT Work starts rolling out on web and mobile to Pro, Enterprise, and Edu users first, with Plus and Business following over the next few days. At the same time, the updated ChatGPT desktop app is available globally on Mac and Windows, with Chat, Work, and Codex now available on every plan, including Free. OpenAI also says the separate Codex app is merging into the new ChatGPT desktop app.

The launch post adds several product details that matter for builders and teams. ChatGPT Work can use plugins to connect to tools such as Slack, Microsoft Teams, Google Drive, SharePoint, email, calendars, and CRMs. OpenAI is also introducing Sites in public beta, plus Scheduled Tasks that let ChatGPT repeat work on a schedule or when events occur.

Why the story is trending on X

The story is trending on X because OpenAI did not frame this as just another model update. The official @OpenAI announcement presented ChatGPT Work as a new agent powered by Codex and GPT-5.6, which immediately pushed the conversation toward what ChatGPT can now do across real workflows instead of what benchmark score moved this week.

That angle travels well on X because it touches several active debates at once: whether agent products can become trustworthy for real work, whether connected apps make these systems more useful or more complicated, and whether coding agents like Codex are now turning into broader workplace agents. For developers and product people, it is a more concrete story than a generic capability claim because the launch is tied to desktop, plugins, browser use, scheduled automation, and document creation in one package.

What this means for developers, builders, or product teams

The practical signal is that OpenAI is trying to collapse several separate surfaces into one work layer. Instead of keeping chat, coding, browser automation, and connected apps as loosely related tools, the company is moving toward a single product where users can start with a goal and let the system produce artifacts, follow-up actions, and recurring workflows.

For product teams, that matters because it raises the bar for what users may expect from assistant software. A chat product that only answers questions looks smaller next to a product that can draft a deck, update a sheet, monitor Slack, and keep working in the background. For developers, the more interesting shift is architectural: the value is no longer just the model itself, but the combination of model, connected tools, permissions, scheduling, and a durable workspace.

What remains unclear

OpenAI has outlined the surface area of ChatGPT Work, but several practical questions remain open. It is still unclear how often users will trust it with truly long-running workflows, how much oversight most teams will want before actions happen, and how predictable usage costs will feel once more people start treating it like a background worker instead of a chatbot.

The rollout is also staggered, which means real-world feedback will probably arrive unevenly across plans. And while OpenAI is positioning ChatGPT Work as a broader work agent, it will take time to see whether users adopt it as a daily operating surface or mostly treat it as a powerful but occasional assistant for bigger tasks.

Sources