ChatGPT Work Bets on Multi-Hour Autonomy, Targeting the Last Moat of Enterprise Software

2. ChatGPT Work Bets on Multi-Hour Autonomy, Targeting the Last Moat of Enterprise Software

OpenAI launched ChatGPT Work on July 7, 2026, combining Codex and GPT-5.6 into a persistent agent that operates across apps and files for multi-hour autonomous sessions. Unlike single-turn assistants, ChatGPT Work accepts a goal, takes action across connected software, and delivers finished output rather than a draft or a recommendation. The product ships with cross-app integrations and is positioned explicitly at knowledge workers running complex, multi-step projects rather than quick queries.

The strategic signal here is the category OpenAI is now contesting directly: Microsoft 365 Copilot and Google Workspace AI. Both have spent 2024-2025 embedding AI into existing productivity suites, betting that distribution and data residency would be the winning moat. ChatGPT Work counters that framing by making the agent itself the workspace. If a user can hand a multi-hour project to ChatGPT Work and receive finished output, the host application matters less. That erodes the integration advantage Microsoft and Google have built around Teams, Outlook, Drive, and Docs. Enterprise IT buyers who justified Copilot on seat-count economics now have a direct performance comparison to make.

The product also marks a line OpenAI has not crossed before: explicit commitment to hours-long autonomous operation. Every prior ChatGPT release, including memory and custom instructions, extended the session but kept the human in the loop at each step. ChatGPT Work removes that assumption. The next pressure point to watch is how Microsoft responds at Ignite 2026 and whether Google accelerates the Gemini-in-Workspace roadmap. Autonomous duration, not model quality, is now the primary competitive axis in enterprise AI.

Source: ChatGPT is now a partner for your most ambitious work