OpenAI has introduced Workspace Agents in ChatGPT, a new feature that allows teams to create shared, persistent AI agents capable of handling complex, long-running workflows. Unlike personal GPTs, these agents are designed for cross-team collaboration, operating within organization-defined permissions and controls.
The Evolution of Teamwork
Workspace Agents are powered by the Codex engine, enabling them to perform tasks that typically require human coordination:
Workflow Automation: Agents can gather context, follow team processes, and keep work moving across multiple tools like Slack, Gmail, Notion, and CRMs.
Proactive Execution: Because they run in the cloud, these agents can operate on a schedule or be deployed in Slack to respond to requests in real-time, even when team members are offline.
Persistent Memory: Agents "learn" from team interactions, improving over time as they gather context and store preferences, making them better at executing specialized team tasks (e.g., month-end accounting or lead qualification).
Built for Enterprise Security
OpenAI has implemented robust governance for Workspace Agents:
Human-in-the-Loop: For sensitive actions like sending emails or adding calendar events, agents can be configured to require explicit human approval.
Admin Controls: Enterprise and Edu admins can manage exactly which tools and data sources user groups can access.
Governance & Analytics: The new Compliance API provides full visibility into agent configuration, updates, and run history. Analytics dashboards allow teams to monitor agent usage and effectiveness.
Getting Started
Users can build an agent in minutes by navigating to the "Agents" tab in the ChatGPT sidebar and describing their workflow. ChatGPT provides a step-by-step wizard to define steps, connect tools, and add "skills." OpenAI has also released a library of templates for common use cases:
Software Reviewer: Routes IT tickets and enforces policies.
Product Feedback Router: Aggregates feedback from Slack/support channels into actionable tickets.
Lead Outreach Agent: Researches and qualifies leads, drafting personalized emails.
Third-Party Risk Manager: Screens vendors against sanctions and reputational risk.
Availability
Preview Status: Available now in research preview for ChatGPT Business, Enterprise, Edu, and Teachers plans.
Pricing: Free until May 6, 2026; credit-based pricing begins thereafter.
Roadmap: Future updates will include automated triggers, better performance dashboards, and deeper integration into the Codex desktop app.
Workspace Agents represent the next step in "Team-as-an-Agent" workflows. By enabling agents to manage cross-functional tasks like code reviews, ticket triaging, and metric reporting, these tools allow engineering teams to offload the "coordination tax" of software development, allowing for faster delivery of high-quality solutions. Organizations can start building their agents today in ChatGPT.
Join the conversation! Your thoughts help the community grow.