OpenAI Launches AI Agents for Teamwork

OpenAI has launched Workspace Agents, a new product designed for enterprises to manage AI workers. Available on ChatGPT Business and higher subscription tiers, these agents can be created or selected from templates to perform tasks across applications such as Slack, Google Drive, and Salesforce.

Powered by Codex, these agents can write and run code, use connected apps, and maintain memory across multiple steps. This capability allows them to handle long-running workflows and operate independently, unlike previous models that paused without user interaction. They can be scheduled for recurring tasks or run on demand.

OpenAI is positioning Workspace Agents as the successor to custom GPTs for organizations. The company is deprecating the custom GPT standard for business users, requiring them to transition to the new agent model. Individual custom GPT users can continue using them for the foreseeable future.

Emphasis is placed on governance, with administrators controlling agent access, permissions, and authentication methods. Write actions default to requiring human approval, with options for customization. OpenAI also claims built-in safeguards against prompt-injection attacks and offers a Compliance API for visibility into agent activity.

Early testers like Rippling report significant time savings, with a sales consultant building an agent that automates account research and call summaries. This launch is part of OpenAI's broader strategy to integrate persistent, connected agents into organizational workflows, competing in a market with similar offerings from Microsoft, Google, and Salesforce.