OpenAI is expanding ChatGPT beyond its chat interface with new "workspace agents." This feature allows business users to automate recurring tasks, even when employees are offline. Powered by the Codex model and designed as "persistent assistants," these agents can connect to third-party applications, retain project context, and execute multistep workflows without constant prompting.
Workspace agents are an evolution of OpenAI's "GPTs," offering no-code customization for specific business functions like report generation, coding, or message response. They run in the cloud and can be shared within an organization, enabling teams to collaborate on and improve agents for use in ChatGPT or Slack.
Users create agents by describing a desired workflow within the ChatGPT interface. The system then maps the process, connects necessary tools, and tests the agent. Once satisfied, users can activate the agent for scheduled runs or specific triggers.
This launch comes amidst intense competition in the AI agent market from major tech players like Google, Microsoft, and Amazon. OpenAI faces pressure from rivals such as Anthropic.
OpenAI is implementing security measures against threats like prompt injection, allowing companies to limit agent data and tool access. Users can also set up approval requirements for sensitive actions.
The workspace agents are currently available as a research preview for ChatGPT Business, Edu, Enterprise, and Teachers subscribers. A credit-based pricing model will be introduced after May 6. Custom GPTs will remain available, with a future mechanism planned for converting them into workspace agents.