OpenAI Launches Workspace Agents in ChatGPT: Codex-Powered Tasks That Run While You're Offline
OpenAI has rolled out Workspace Agents in ChatGPT, a significant upgrade to its enterprise product that turns the assistant into a platform for long-running, shared automation. The launch is in research preview across ChatGPT Business, Enterprise, Edu, and Teachers plans.
Unlike standard GPTs, Workspace Agents are powered by Codex and run in the cloud continuously — they do not require a user to be present in a chat session to make progress. Teams build an agent once, share it across the organisation, and can deploy it inside ChatGPT or Slack without exporting it.
What They Can Do
OpenAI described five agent types that its own sales and engineering teams are already using internally:
- Software Reviewer: Reviews employee software requests, checks them against approved tools and security policies, files IT tickets automatically.
- Product Feedback Router: Monitors Slack, support channels, and public forums; converts feedback into prioritised tickets and weekly product summaries.
- Weekly Metrics Reporter: Pulls data every Friday, builds charts, writes the narrative summary, and sends the report without prompting.
- Lead Outreach Agent: Researches inbound leads, scores them, drafts personalised follow-up emails, updates the CRM.
- Third-Party Risk Manager: Assesses vendor risk including sanctions exposure, financial health, and reputational signals; outputs a structured report.
Workspace Agents can connect to dozens of enterprise tools for context and actions. OpenAI says a team describes a workflow, and ChatGPT turns it into a working agent with defined steps, connected tools, and a test cycle before deployment. Templates are available for finance, sales, and marketing to speed setup.
GPTs Are Not Immediately Removed
OpenAI has been careful not to immediately retire GPTs. Both products will run in parallel while enterprise customers test Workspace Agents against their existing workflows. A migration path to convert existing GPTs into Workspace Agents is coming.
Competitive Context
The launch is a direct push into territory where Anthropic has spent the last three months building: Claude Managed Agents ($0.08/hr) and Claude for Microsoft Word position Anthropic as the default for knowledge worker automation. OpenAI’s response is a platform play — embedding agents into the interfaces enterprise teams already use (Slack, CRM, IT ticketing) rather than requiring a separate tool to manage them. The emphasis on shared team agents, not individual assistants, signals that OpenAI sees organisational workflow ownership as the enterprise battleground in 2026.