OpenAI wants to own your workday | Friday, July 10, 2026

It merged ChatGPT, its Codex coding app, and the Atlas browser into a single super app and launched ChatGPT Work, an agent powered by GPT-5.6 that gathers context across your files and apps and returns finished documents, spreadsheets, presentations, and working web apps.

The reframing matters more than any feature. OpenAI is no longer competing for your query. It is competing for your seat, the place work actually happens, which puts it head to head first with Anthropic's Claude Cowork, the rival agent it was built to answer, and then with the incumbent suites, Microsoft 365 Copilot and Google Workspace. And it is doing it ahead of a planned IPO, which tells you this is about durable revenue, not product elegance.

The strategy is a land grab for the workday, and the first target is Anthropic. ChatGPT Work is OpenAI's direct answer to Claude Cowork, the general knowledge-work agent Anthropic launched in January and pushed to web and mobile just two days earlier. Both companies built these agents out of their coding tools, and both now aim them at the same prize: the ordinary business work, the memos, spreadsheets, decks, and reports that Anthropic's own usage data says is more than 90% of what Cowork already does. Behind that duel sit the incumbents whose desktops are the ultimate territory, Microsoft 365 Copilot and Google Workspace. Owning the seat, not the query, is where the durable per-user revenue lives, and it is not a coincidence that this arrives as OpenAI readies an IPO.

For enterprise leaders, this is a buy-decision dressed as a feature update. ChatGPT Work is genuinely useful if your teams keep bouncing between a chatbot, a coding tool, and their real applications, and an agent that ships finished artifacts collapses that. But adopting it means standardizing your most sensitive workflows on one vendor's surface, with your data, context, and habits accumulating inside it. That is the deepest kind of lock-in, and it is being offered right as the model layer underneath commoditizes.

The pitch is seductive: one surface, four tools collapsed, an agent that ships finished work. The catch is what you standardize when you say yes. ChatGPT Work accrues your data, your context, your team's habits, and your most sensitive workflows inside one vendor's environment, and the leverage it delivers is exactly the thing that makes leaving expensive. This is arriving at the precise moment the model underneath stopped being a moat. Grok 4.5 reached the frontier's neighborhood at a third of the price, and even Google's flagship slipped six weeks. Capability is commoditizing. The surface is where lock-in migrates.

That split is the enterprise buyer's whole decision, and it holds whichever way the race goes. The model is now a configuration choice you can swap; the workday surface is an architectural commitment you cannot. Adopting either super app, ChatGPT Work or Claude Cowork, for real productivity gains is defensible. Letting whichever you pick quietly become the system of record for how your company works, with no portability plan and no second surface, is how you wake up in a year with your leverage gone and your renewal priced accordingly. Take the productivity. Keep the exit.

That is a bigger idea than a feature, and it is worth sitting with. The surface where work happens has been contested territory for decades, owned first by the operating system, then the browser, then the productivity suite. OpenAI is making a claim on it, Anthropic is already there with Claude Cowork, and Microsoft and Google will not concede it quietly. For once the interesting question for an enterprise buyer is not which model is best. The models are converging and getting cheaper by the month. The interesting question is whose surface your people will actually live inside, because that choice is stickier than any model decision and far harder to reverse.