OpenAI rolled out AI agents in ChatGPT Workspace on April 23, 2026, giving businesses, schools, and enterprises autonomous cloud‑based assistants.

Driving the news: OpenAI says the new agents replace the older Custom GPTs and run nonstop in the cloud, so tasks keep moving forward even after you close the app.

Why it matters: Switching to always‑on agents lets teams automate long‑running workflows (code generation, report drafting, multi‑step messaging) without having to click a button each time. With processing moved off the user's device, OpenAI hopes to boost productivity while keeping corporate data private.

The cloud handles persistence while users maintain control over their workflows.

What they're saying: Company spokespeople described the feature as "an evolution toward fully autonomous AI collaborators." Analysts note that the move aligns OpenAI with rivals offering server‑side AI services. In our earlier coverage of OpenAI's desktop superapp combining ChatGPT, Atlas, and Codex, we highlighted the company's push to bundle multiple AI tools under one interface, a strategy now extended to the Workspace agents.

What's next: OpenAI plans to let users convert existing Custom GPTs into agents and to broaden beta access to additional subscription tiers. Future updates may add richer integrations with third‑party platforms, further embedding autonomous AI into everyday business tools.

The question isn't whether agents will automate work. It's whether they'll do so in ways we can inspect, correct, and trust.