OpenAI Launches Workspace Agents to Automate Enterprise Workflows

OpenAI began rolling out Workspace Agents in ChatGPT on April 22, giving enterprise users AI that can run complex, multi-step tasks across a company's existing software without waiting for a human to initiate each step. The product is in research preview and available on the Business, Enterprise, Education, and Teachers tiers.

The practical scope is fairly broad. Workspace Agents can draft and send messages in Slack, retrieve documents from Google Drive, update Salesforce records, and work across Notion, Atlassian, and Microsoft Teams and Excel. They keep going until a task is complete, which is a real shift from the current model of ChatGPT as a tool you consult rather than one that executes on your behalf.

Pricing is free through May 6, after which OpenAI moves to credit-based billing. Specific rates have not yet been disclosed.

OpenAI frames the product as the successor to Custom GPTs, which let users build specialised chatbots but stopped short of autonomous action. Workspace Agents take that further: they do not just answer questions, they do things. The rollout includes admin governance controls that let organisations decide which tools their agents can access and what changes they are permitted to make.

The strategic logic is fairly clear. OpenAI is competing for enterprise contracts against Anthropic's Claude, Microsoft Copilot, and Google's Gemini. Sora's winding down and the framing of GPT-5.5 as a super-app foundation both point in the same direction: OpenAI is betting its commercial future on becoming the operating layer for how businesses get work done. Workspace Agents are what that looks like in practice.

Whether organisations are willing to hand that level of autonomy to AI agents is an open question. The credit-based pricing that kicks in on May 7 will start to answer it.