OpenAI has introduced workspace agents, a major upgrade to its custom GPT feature that lets teams build and share AI assistants powered by Codex technology. These agents can handle complex, long-running tasks like report preparation, code writing, and message responses while operating within organizational permissions and controls. The feature is now available as a research preview through ChatGPT Business, Enterprise, Edu, and Teachers plans.

What Are Workspace Agents and How Do They Differ From Custom GPTs?

Workspace agents represent a significant evolution from the custom GPTs that OpenAI introduced last year. While custom GPTs were designed for individual use and never gained major traction within ChatGPT, workspace agents are purpose-built for team collaboration. The key difference lies in their architecture and deployment model. Workspace agents run in the cloud, meaning they can continue working on tasks even when team members are offline. They're also designed to be shared across an entire organization, allowing multiple people to use the same agent and improve it over time.

The technology powering these agents is Codex, OpenAI's specialized model for understanding and generating code. This foundation gives workspace agents the ability to handle technical tasks alongside more general business workflows. Teams can deploy a single agent across ChatGPT and Slack, creating a unified experience for their organization.

What Specific Tasks Can Workspace Agents Handle?

OpenAI has outlined several practical applications for workspace agents in team environments. These capabilities span multiple business functions and suggest how the technology could reshape daily workflows:

• Report Preparation: Agents can gather data, format information, and generate reports without manual compilation by team members.
• Code Writing and Development: Powered by Codex, agents can write, review, and suggest code improvements for technical teams.
• Message Responses: Agents can handle routine communications, freeing team members from repetitive messaging tasks.
• Complex Task Automation: Long-running workflows that previously required manual oversight can now run autonomously within organizational guardrails.

How to Deploy Workspace Agents in Your Organization

For teams interested in testing this new capability, the process involves several key steps. Organizations with ChatGPT Business, Enterprise, Edu, or Teachers plans can access workspace agents immediately as a research preview. OpenAI has also announced plans to allow direct conversion of existing custom GPTs into workspace agents, making migration straightforward for teams that have already built custom tools.

• Access Requirements: Your organization must have an active ChatGPT Business, Enterprise, Edu, or Teachers subscription to use workspace agents.
• Agent Creation: Teams can create shared agents tailored to their specific workflows and organizational needs.
• Permission Management: Agents operate within permissions and controls set by your organization, ensuring security and compliance.
• Integration Options: Agents work within ChatGPT and can be integrated with Slack for seamless team access.
• Continuous Improvement: Teams can refine and improve agents over time based on real-world usage and feedback.

Why This Matters for Enterprise Teams

The introduction of workspace agents addresses a real gap in how organizations use AI tools. Previous AI features were often designed for individual productivity, leaving teams to figure out how to collaborate around them. Workspace agents flip this model by treating team collaboration as the primary use case from the start. The cloud-based architecture means that complex, time-consuming tasks can run continuously, completing work even outside business hours. This is particularly valuable for global teams operating across multiple time zones.

The integration with Slack is also significant. Slack has become the central hub for team communication in many organizations, so embedding AI agents directly into that workflow reduces friction and increases adoption. Rather than asking teams to switch between tools, workspace agents meet them where they already work.

OpenAI's announcement of workspace agents comes as part of a broader launch week that included updates to ChatGPT Images 2, a new version of Codex for Mac, and the introduction of Chronicle in Codex. This concentrated release suggests OpenAI is making a strategic push to position ChatGPT as an enterprise-grade platform rather than just a consumer tool.

The research preview status means organizations can test workspace agents now, but the feature may still undergo changes before a full release. Teams interested in early access should explore the feature through their existing ChatGPT subscriptions and provide feedback to OpenAI about how well the agents meet their specific needs.