OpenAI introduced workspace agents in ChatGPT on April 22, 2026, enabling teams to create shared agents that handle complex tasks and long-running workflows within organizational permissions and controls. The workspace agents ChatGPT feature represents an evolution of GPTs, powered by Codex technology that allows agents to take on tasks ranging from report preparation to code writing to message responses.

Workspace agents operate in the cloud and continue functioning even when users are offline. Teams can build an agent once, deploy it across ChatGPT or Slack, and refine it over time based on usage patterns. The agents gather context from multiple systems, follow team processes, request approval when necessary, and maintain workflow continuity across different tools.

How Workspace Agents Function

The workspace agents ChatGPT system enables agents to write or run code, use connected applications, retain learned information, and continue work across multiple steps. Users can set agents to run on schedules or deploy them in Slack to handle requests as they arrive. For example, OpenAI’s sales team uses an agent to compile details from call notes and account research, qualify new leads, and draft follow-up emails directly in representatives’ inboxes.

Getting started requires clicking the Agents option in the ChatGPT sidebar and describing a workflow the team performs regularly. ChatGPT guides users step by step through converting the workflow into an agent. The platform helps define steps, connect appropriate tools, add skills, and test functionality until it meets expectations.

Five Agent Templates for Teams

  • Software Review Agent: Triages software requests, enforces policy, routes approvals, and opens IT tickets with clear next steps
  • Product Feedback Routing Agent: Captures feedback from Slack, support, and public channels, prioritizes issues, and generates weekly product action items
  • Weekly Metrics Reporting Agent: Auto-pulls Friday data, generates charts, drafts narratives, and delivers business reports
  • Lead Outreach Agent: Qualifies inbound leads, drafts tailored follow-ups, and updates CRM systems
  • Third-Party Risk Management Agent: Screens vendors for sanctions, financial, and reputational risk, then delivers structured reports

Teams can also access templates for finance, sales, marketing, and other functions. Each template includes built-in skills and suggested tools, allowing rapid setup and customization. OpenAI’s accounting team built an agent that prepares key month-end close components, from journal entries to balance sheet reconciliations to variance analysis, completing work in minutes while generating workpapers with underlying inputs and control totals.

Enterprise Controls and Governance

Workspace agents include enterprise-grade monitoring and controls for organizational protection. ChatGPT Enterprise and Edu administrators can control which connected tools and actions user groups access. Admins manage who has permission to use, build, and share agents. For sensitive steps like editing spreadsheets, sending emails, or adding calendar events, agents can be configured to request approval before proceeding.

Built-in safeguards help agents remain aligned with organizational instructions when encountering misleading external content, including prompt injection attacks. The Compliance API provides administrators visibility into every agent’s configuration, updates, and runs, enabling monitoring and control of how agents are built and used. Admins can suspend agents if needed. Analytics help teams see how live workspace agents are being used, including completion counts and user numbers.

Availability and Pricing

Workspace agents are available in research preview for ChatGPT Business, Enterprise, Edu, and Teachers plans. For Enterprise and Edu plans, administrators can enable agents using role-based controls. The feature will be free until May 6, 2026, with credit-based pricing beginning on that date. OpenAI plans to add new triggers for automatic work initiation, improved dashboards for performance understanding, expanded action capabilities across business tools, and support for workspace agents in the Codex app.

Early testers report more consistent results and increased time for higher-value work. Rippling’s AI Engineering team noted that one Sales Consultant built, evaluated, and iterated a Sales Opportunity agent without engineering support. The agent researches accounts, summarizes call recordings, and posts deal briefs directly to team Slack rooms, automating work that previously required 5-6 hours weekly per representative.

Source: OpenAI