Workspace agents in ChatGPT enable teams to create shared, cloud-powered assistants that handle complex tasks and long-running workflows within organizational permissions and controls. These agents, powered by Codex, automate common work activities such as report preparation, coding, and message responses. Designed for team use, they can be built once and shared across ChatGPT or Slack, improving over time through interaction and feedback.
Key Features and Use Cases
Workspace agents gather context from various systems, follow team processes, request approvals when necessary, and maintain workflow continuity across tools. For example, OpenAI’s sales team uses an agent to compile call notes, qualify leads, and draft follow-up emails, reducing manual effort and increasing customer engagement.
Teams can create agents by describing workflows in ChatGPT’s Agents sidebar, with guided step-by-step setup. Workspace agents are currently available in research preview for ChatGPT Business, Enterprise, Edu, and Teachers plans. GPTs remain available during this transition, with future options to convert GPTs into workspace agents.
Example Agents Teams Can Build
- Software Reviewer: Reviews software requests, enforces policies, routes approvals, and opens IT tickets.
- Product Feedback Router: Collects feedback from Slack and support channels, prioritizes issues, and generates product action summaries.
- Weekly Metrics Reporter: Automatically pulls data, creates charts, drafts narratives, and shares reports.
- Lead Outreach Agent: Qualifies leads, drafts personalized follow-ups, and updates CRM systems.
- Third-Party Risk Manager: Screens vendors for sanctions, financial, and reputational risks, producing structured reports.
Templates are available for finance, sales, marketing, and more, with built-in skills and suggested tools for quick customization.
Integration and Continuous Operation
Workspace agents operate in the cloud with access to files, code, tools, and memory, enabling multi-step workflows and continuous operation even when users are away. They can run on schedules or respond to requests in Slack, such as answering employee questions, linking documentation, and filing tickets to prevent workflow blockages.
Currently, agents interact via ChatGPT and Slack, with plans to expand integration into additional platforms, allowing them to assist directly within existing team workflows.
Knowledge Sharing and Improvement
Agents help consolidate scattered knowledge into reusable workflows that adhere to organizational processes and policies. For instance, OpenAI’s accounting team uses an agent to automate month-end close tasks, generating workpapers and ensuring compliance. Agents improve over time through memory and conversational corrections, enabling teams to maintain up-to-date workflows and share or duplicate agents for new tasks.
Control, Security, and Governance
Users retain control over agents by specifying accessible tools, data, and required approvals for sensitive actions like sending emails or editing documents. Analytics provide insights into agent usage and performance.
Enterprise-grade monitoring and controls allow admins to protect sensitive data, manage user permissions, and oversee agent creation and deployment. Built-in safeguards defend against misleading content and prompt injection attacks. The Compliance API offers visibility into agent configurations and activity, with capabilities to suspend agents if necessary. Upcoming features will enable admins to view all organizational agents, usage patterns, and connected data sources from a centralized console.
Summary
Workspace agents extend ChatGPT’s capabilities by enabling teams to automate and share complex workflows securely and efficiently. They integrate deeply with organizational tools, support continuous operation, and improve through use, all while maintaining strict governance and control for enterprise environments.
