OpenAI chief technology officer Mira Murati said on Tuesday that the company is expanding its ChatGPT platform with new workspace agents designed to handle complex tasks in the cloud. These agents run on Codex models and can perform multi-step workflows across different software tools while maintaining security standards. The announcement came during a keynote at the company’s annual developer conference in San Francisco.
The agents operate as background processes in ChatGPT, allowing teams to delegate repetitive or technical assignments without manual oversight. They can execute sequences such as data extraction from spreadsheets, automated report generation, or updating project management systems. All operations occur within OpenAI’s cloud infrastructure, which the company says meets enterprise-grade security requirements.
Murati emphasized that these agents are not replacements for human workers but tools to reduce routine workload. Early access partners include several Fortune 500 companies in finance and logistics. She added that feedback from these partners shaped the agents’ design, particularly around security and audit trails.
The move reflects OpenAI’s push to embed its models deeper into business software ecosystems. Earlier this year, the company introduced plugins that let ChatGPT connect to third-party apps. Workspace agents extend this capability by performing actions directly, rather than just providing suggestions. The feature is available now to enterprise customers on paid ChatGPT tiers.
Source: openai.com