OpenAI adds admin controls to ChatGPT Business

ChatGPT Business is getting more like real workplace software, not just a chatbot people expense on the side. That is the point of OpenAI’s latest update. On April 22, the company added workspace agents for Business customers, plus tighter admin controls for connected apps, scheduling, analytics, and Slack governance. ### What actually shipped? The biggest new piece is workspace agents. Teams can build agents from templates or from scratch, connect them to tools like Google Drive, Google Calendar, Slack, and SharePoint, add files or custom MCP servers, and then run those agents inside ChatGPT or in Slack. They can also share agents privately, by link, or through a workspace directory, and schedule recurring runs for repeatable work. ### Why do admin controls matter here? Because a business chatbot stops being a toy the second it can touch internal systems. OpenAI now lets workspace admins manage access to agent building, publishing, and Slack usage through admin controls. On the app side, admins can decide whether an app gets all actions, only read actions, or a custom set of actions. That is based on your tools.” ### What changed for connectors? The naming changed first. OpenAI renamed connectors to “apps” in December 2025, but kept the underlying behavior. So when the release notes talk about connected tools and app controls, this is the same general product surface getting folded into a cleaner enterprise admin model. Existing connectors and company knowledge setups keep working. ### How granular are these controls? More granular than the old “on or off” story. In ChatGPT Business, admins can control which apps are enabled for the workspace and manage role-scoped permissions. If a new app is enabled, admins can assign which roles may use it. If that app supports actions, admins can review those actions before rollout. And if the app later adds new actions, admins can choose allowed by default, or blocked until review. That is the kind of boring feature enterprises care about a lot. ### What makes this feel more enterprise? Two things — auditability and visibility. Workspace agents now include version history and analytics, which means teams can track changes and usage instead of treating prompts like disposable chat logs. OpenAI also refreshed Workspace analytics for Business on April 16, replacing user-level analytics with a workspace-level dashboard that shows active users, a view of adoption and cost. ### Is this totally new for ChatGPT Business? Not exactly. OpenAI had already been moving this way. In June 2025, it pitched business-plan updates around connectors to internal tools, custom connectors with MCP, and new security controls. The new release looks like the next step — less about adding one flashy capability, more about making the whole setup governable enough for broader deployment. ### What is the catch? Business still is not identical to Enterprise. The app-control article makes clear that Enterprise and Edu start with all apps disabled by default, while Business uses workspace admin controls and role-scoped permissions in a somewhat lighter model. So OpenAI is narrowing the gap, but not erasing plan differences. The conversation has shifted. It is no longer just “is the model good?” It is “can IT approve this without losing control?” OpenAI’s update is an answer to that. The company is turning ChatGPT Business into something procurement, security, and team leads can actually deploy — with permissions, logs, and guardrails instead of vibes.