OpenAI spent years building the world’s most widely used AI assistant. On July 9, 2026, they announced what comes next: ChatGPT Work, a full enterprise AI platform where agents don’t just answer questions — they own workflows, run in the background, and get shared across entire teams.

The livestream, billed as “the next chapter for ChatGPT,” bundled three distinct releases: a new ChatGPT Work agent, an upgraded desktop app with Codex built in, and a hosted sites service for creating AI-powered web experiences. Taken together, it marks a clear shift from AI-as-assistant to AI-as-employee.

What OpenAI Shipped on July 9

ChatGPT Work agent. Workspace agents can now take on complete, multi-step workflows — not just single tasks. They can prepare reports, write and execute code, respond to messages, and follow team-defined processes. Critically, they run in the cloud, meaning they keep working after you close your laptop. Teams build an agent once and share it across the whole organisation, so the same trained workflow is available to everyone with access.

Codex inside the desktop app. The updated macOS desktop app now ships with Codex integrated. Users get a project sidebar, a thread list, and a review pane for tracking parallel agent threads across multiple long-running tasks. The app also pulls in GitHub pull requests directly, letting developers review code and agent-generated diffs side by side without switching tools. Users can edit Markdown and code inline and ask Codex to revise selected sections.

Hosted sites. A new hosted sites feature lets paid users publish AI-powered web pages directly from ChatGPT. Enterprise customers can create internal dashboards, reports, or lightweight apps without a separate deployment pipeline.

ChatGPT for PowerPoint. Generally available for Enterprise customers, the PowerPoint integration lets teams create and revise editable decks inside Microsoft PowerPoint, ask questions about narrative and slide structure, and generate slides from reusable Skills and connected apps. It remains free for Enterprise through August 6, after which token-based pricing applies.

Model Access

ChatGPT Work runs on the newly released GPT-5.6 family:

• Free and Go users — access GPT-5.6 Terra
• Plus, Pro, Business, Enterprise — choose from GPT-5.6 Sol, Terra, or Luna, and set the effort level per task
• Pro and Enterprise — unlock Ultra mode, the highest reasoning tier, for the most complex tasks

Why This Matters for Enterprise Teams

The shift here is structural. Earlier iterations of workspace agents were useful for single-person productivity — one person getting better answers, faster. ChatGPT Work is positioned for team-scale deployment: shared agents that carry team knowledge, follow team processes, and run unsupervised on workflows that used to require a human in the loop.

The move to token-based pricing for workspace agent runs (which started July 6) also signals that OpenAI sees this as a consumption-driven enterprise product, not a flat-fee subscription feature. Teams that deploy agents heavily will pay more — but in exchange for measurable, automated output.

The Codex integration is the other significant tell. Codex has been positioned as a software engineering agent since its relaunch in 2025. Bundling it into the general ChatGPT desktop app means OpenAI is blurring the line between its developer tools and its business tools. Non-technical teams can now trigger code-writing agents from the same interface they use for writing and research.

What This Means for Business

If you’re evaluating AI for your business right now, this release changes the comparison. ChatGPT Work is no longer just competing with Copilot for individual productivity — it’s competing with purpose-built workflow automation tools. The question isn’t “should we use AI?” It’s “which AI platform should own our workflows?”

If you’re already running AI pilots, the agent-sharing model is worth attention. The ability to build a workflow agent once and share it across a team dramatically lowers the deployment cost per use case. A well-designed agent for invoice processing, client reporting, or meeting prep becomes a team asset, not an individual tool.

If you’re a data team, the Codex integration in the desktop app matters. Parallel agent threads, GitHub PR review in the sidebar, and inline code editing mean that Codex is now a first-class option for teams doing data analysis, model deployment, and pipeline work who want an AI agent embedded in their workflow — not running in a separate tool.

The timing is notable. July 9 also saw SpaceXAI launch Grok 4.5, and GPT-5.6 Sol, Terra, and Luna go fully public. OpenAI didn’t lead with the model launch — they led with the work platform. That’s a deliberate signal about where the company sees the next competitive battleground: not model benchmarks, but enterprise deployment.

The Bigger Picture

The enterprise AI market is rapidly consolidating around a few large platforms — Microsoft Copilot, Google Workspace AI, Salesforce Agentforce, and now ChatGPT Work. Each is betting that companies will anchor on a single AI work surface rather than stitching together point solutions.

For smaller businesses and mid-market companies, this is genuinely useful news. These platforms are built for Fortune 500 IT budgets in terms of integrations and governance, but token-based pricing means smaller teams can access the same core capability at lower volumes.

For Enterprise DNA clients exploring AI deployment, the key question remains the same regardless of which platform: where is the highest-value process you can replace with an agent that runs unsupervised? That question matters more than which AI brand powers it.

---

Enterprise DNA builds custom AI agent deployments for businesses that need more than an off-the-shelf platform. If you’re evaluating ChatGPT Work, Copilot, or custom AI agents for your operations, book a discovery call to map the right approach for your use case.