OpenAI launched ChatGPT Work on Thursday, a workspace that pairs its chatbot with Codex, its coding tool, to turn a team’s scattered files into finished documents, spreadsheets, presentations, and websites. The product pulls context from a company’s connected tools and acts directly across files and desktop applications rather than only answering questions about them.
The launch matters because OpenAI is aiming the tool at paid seats inside companies, not the free chat window most consumers know. Enterprise contracts carry higher margins and stickier renewal cycles than individual subscriptions, and ChatGPT Work is OpenAI’s clearest bid yet to own that line item in a corporate software budget.
ChatGPT Work is also OpenAI’s most direct answer to Claude Cowork, the agent Anthropic released in January that plans and executes multi-step tasks autonomously across a user’s software. Both companies are now selling the same pitch to the same buyer: an AI system that finishes the work rather than drafts an answer for a human to finish.
Access to ChatGPT Work began Thursday on web and mobile for Pro, Enterprise, and Edu subscribers. OpenAI said Plus and Business users will get access over the following days, a staged rollout that mirrors how the company typically tests enterprise features on its highest-paying tiers before opening them more broadly.
The tool runs on GPT-5.6, the model family OpenAI made publicly available the same week. Pairing a new flagship model with a new product category lets OpenAI market both releases as a single event rather than two separate announcements competing for the same attention.
OpenAI’s announcement does not include independent benchmark results comparing ChatGPT Work’s document and spreadsheet output to Claude Cowork’s, or to the manual workflows both are meant to replace. Whether an agent that produces a finished slide deck actually saves a team time depends on how much rework the output needs, a detail neither company has published.
Google made a similar promise when it built Gemini into Workspace in 2024, describing it as a layer that would generate first drafts of documents and spreadsheets from a prompt. Verifiable, independent adoption figures for that feature remain scarce two years later. That pattern suggests document-generation features are easier to ship than to prove valuable.
The stakes differ by buyer. For OpenAI and Anthropic, enterprise seats are the more durable revenue base than consumer subscriptions, so the ChatGPT Work and Claude Cowork rivalry will likely be decided by procurement teams and pilot programs, not by app store rankings. For consumers, neither release changes what the free or Plus chat experience looks like today.
Enterprise buyers currently evaluating Claude Cowork should treat ChatGPT Work as a genuine second bid rather than a follow-on feature, since both products now compete on the identical claim of finished work product rather than chat responses. Teams running pilots of either tool should measure how much manual editing the output actually requires before committing to a seat-based contract at scale.
Based on OpenAI’s July 9, 2026 announcement introducing ChatGPT Work.
