July 10, 2026, (Inside AI) — OpenAI today launched ChatGPT Work, a new agentic capability that transforms ChatGPT from a conversational assistant into an autonomous project partner. The feature, powered by the new GPT‑5.6 frontier model, can independently gather information across apps, create finished materials like sheets and slides, and manage complex, multi-step workflows for hours without supervision.
ChatGPT Work is rolling out immediately on web and mobile for Pro, Enterprise, and Edu plans, with Plus and Business users gaining access in the coming days. The updated desktop app, available globally on Windows and Mac, now includes Chat, Work, and Codex for all plans, including Free. This release marks a strategic shift for OpenAI, embedding agentic AI directly into its flagship product and merging the previously standalone Codex coding agent into the main ChatGPT experience.
The launch comes as enterprise AI adoption reaches an inflection point. While competitors like Anthropic and Google have introduced autonomous agents, OpenAI’s move leverages its massive user base—over 5 million weekly Codex users—to mainstream agentic workflows. The company claims nearly 100% of its internal teams now use ChatGPT Work and Codex, signaling confidence in the technology’s readiness for real-world tasks.
From Chat to Autonomous Execution
ChatGPT Work is designed to handle tasks that extend beyond simple Q&A. Users can ask it to analyze a month-end budget variance, transform source materials into a marketing campaign brief, or prepare for a sales meeting. The agent can follow progress, ask clarifying questions, and wait for approval before taking important actions. Crucially, it can chain entire workflows: turning customer research into a brief, then creating marketing assets, and finally adapting those assets for different markets—all while maintaining context across steps.
A standout feature is Scheduled Tasks, which allows ChatGPT Work to operate even when the user is offline. For example, it can monitor Microsoft Teams and Slack messages, automatically update documents or slides based on new information, and share changes with team members. This persistent, asynchronous capability positions ChatGPT Work as a virtual team member rather than a passive tool.
The underlying GPT‑5.6 model is key. OpenAI states it makes ChatGPT “state of the art at reasoning through multi-step tasks and creating materials that follow your templates and reference files.” This model upgrade, rolling out alongside ChatGPT Work, reportedly enhances the system’s ability to handle complex, context-heavy projects—a critical requirement for enterprise adoption where accuracy and consistency are paramount.
Desktop Power and Codex Convergence
On desktop, ChatGPT gains new capabilities that blur the line between cloud and local computing. A built-in browser lets it research markets, compare sources, and pull information from websites or online files within Google Workspace and Microsoft 365. More ambitiously, Computer Use enables ChatGPT to control the user’s machine—clicking, typing, and moving files across apps—to execute tasks in the background. This can be a one-time action or part of a Scheduled Task, effectively automating repetitive desktop workflows.
The Codex app is now fully merged into the ChatGPT desktop app. Developers retain the same powerful coding agent, now enhanced with inline editing within diffs, pull request review in a side panel, faster computer use powered by GPT‑5.6, and support for multiple repositories in a single project. Existing Codex users can update their app to transition seamlessly, with the option to set Codex as the default view and keep its logo. The old ChatGPT desktop app will be renamed ChatGPT Classic.
This integration follows a broader industry trend of consolidating AI tools into unified platforms. Microsoft’s Copilot and Google’s Duet AI similarly embed agents across productivity suites, but OpenAI’s approach is more open-ended, relying on a plugin ecosystem to connect with third-party apps like Slack, Salesforce, and SharePoint. A new unified plugins directory and the ability to invoke specific apps with an “@” command streamline setup.
OpenAI is also launching Sites in public beta, allowing users to turn work into interactive web apps or dashboards shareable via URL. These can be live project trackers, internal portals, or prototypes that ChatGPT updates as underlying data changes. This feature directly competes with no-code platforms like Notion and Airtable, but with AI-driven dynamism.
Security remains a focal point. Enterprise and Edu admins can centrally manage access, tool connections, and action permissions. A new Auto-review system uses advanced models to scrutinize important actions before execution, blocking unauthorized data sharing. OpenAI claims that during adversarial red teaming, auto-review stopped 100% of attempts to extract protected data, including novel attacks. The Compliance API provides visibility into conversations and actions at scale, addressing enterprise oversight needs.
Usage for ChatGPT Work deviates from standard chat metrics. Complex tasks consume more of a plan’s included usage, following a structure similar to Codex. Admins can set spend controls, group limits, and individual overrides to manage costs as adoption grows. This granularity is essential for organizations wary of unpredictable AI expenses, a pain point highlighted by early enterprise AI deployments.
Despite the fanfare, questions linger. Agentic AI systems have historically struggled with edge cases and context drift in long-running tasks. OpenAI’s internal testing may not fully reflect the chaos of real-world enterprise environments. Competitors like Adept and Cohere are also racing to deliver reliable agents, and the market is still defining what “success” looks like beyond demos. The sunsetting of the standalone Atlas browser—an earlier agentic experiment—suggests that OpenAI is still iterating on the user experience.
For now, ChatGPT Work represents a significant bet that the future of AI lies not in answering questions, but in doing work. As OpenAI puts it, this is a first step toward “helping everyone turn their biggest ideas into reality.” Whether it can deliver on that promise without stumbling over the messy realities of daily work remains to be seen.
