OpenAI workspace agents roll out
- OpenAI said on April 22 it is rolling out workspace agents in ChatGPT, shared Codex-powered assistants teams can build, publish, and run across workplace tools. - The agents run in the cloud, can be scheduled or triggered, and work inside ChatGPT or Slack with apps like Google Drive, Calendar, and SharePoint. - The launch replaces one-off custom GPT use with shared, governed automation for paid workspaces. (openai.com)
OpenAI said on April 22 that ChatGPT workspaces can now create shared “workspace agents” that handle repeatable team tasks across connected apps. (openai.com) These agents are powered by Codex, run in the cloud, and keep working after a user closes ChatGPT. OpenAI said teams can use them for reports, coding, message responses, and other long-running workflows. (openai.com) OpenAI’s help pages say agents can be built from templates or from scratch, previewed before publishing, shared inside a workspace, and run on a schedule. They can operate in ChatGPT or Slack. (help.openai.com 1) (help.openai.com 2) The basic idea is to turn a chat assistant into a reusable worker for a team. Instead of one person repeating the same prompt, a workspace can publish one agent with the same tools, files, and instructions for everyone. (openai.com 1) (openai.com 2) That changes how ChatGPT is packaged for office use. OpenAI called workspace agents “an evolution of GPTs,” shifting from single-user custom bots to shared agents that operate inside organizational permissions and controls. (openai.com) OpenAI’s business page says admins can govern which tools agents use, while release notes say eligible workspaces can connect Google Drive, Google Calendar, Slack, SharePoint, files, skills, and custom Model Context Protocol servers. (openai.com) (help.openai.com) The rollout is a research preview for ChatGPT Business, Enterprise, Edu, and Teachers plans, according to OpenAI’s product post and developer cookbook. That limits the launch to paid organizational tiers rather than consumer ChatGPT accounts. (openai.com) (developers.openai.com) OpenAI’s academy materials draw a line between deterministic automation and agents that make bounded decisions from context. In practice, that means more flexibility than a fixed workflow, but also more need for review, permissions, and clear stop rules. (openai.com) The immediate pitch is simple: build one agent once, share it with a team, and let it run the same process in ChatGPT or Slack without rebuilding the workflow every time. (openai.com)
