OpenAI Launches Workspace Agents: ChatGPT Becomes a Full Team Automation Platform

What Just Changed in ChatGPT

On April 22, 2026, OpenAI announced Workspace Agents in ChatGPT — a feature the company describes as "an evolution of GPTs." This is not a minor iteration. Where Custom GPTs were essentially a way to configure a chatbot for a specific task, Workspace Agents are autonomous cloud-based systems that operate continuously, handle multi-step workflows, and collaborate across entire teams. ChatGPT is no longer just a conversational assistant; it is now positioning itself as a full workplace automation platform.

The feature is currently available as a research preview for ChatGPT Business, Enterprise, Edu, and Teachers plans, with free access through May 6, 2026, after which credit-based pricing begins.

Key Features and Capabilities

1. Persistent Cloud Operation

Unlike standard ChatGPT interactions that exist only within a session, Workspace Agents run persistently in the cloud. They continue working even when users are offline. This enables genuinely autonomous workflows — an agent can be assigned a task at the end of the day and return completed work by morning.

2. Codex-Powered Execution

Each agent is powered by OpenAI's Codex technology, giving it the ability to write and execute code, interact with connected applications, and manage multi-step technical workflows. This is a substantial upgrade from the text-completion-focused Custom GPTs.

3. Shared Team Context

Workspace Agents are designed for organizational use. A team creates an agent once, and every member can use it within ChatGPT or Slack. Agents retain memory across sessions, pull context from connected tools, and can be collaboratively refined over time — addressing one of the core limitations of previous GPT customization tools.

4. Slack Integration and Scheduling

Agents can be embedded directly into Slack channels, where they automatically detect and act on relevant requests without manual triggering. Teams can also put agents on a schedule for recurring tasks such as weekly reporting, data summaries, or codebase health checks.

5. Approval Workflows for Sensitive Actions

For actions with real-world consequences — sending emails, creating calendar entries, executing transactions — agents are designed to pause and request human approval. This builds a layer of oversight into automated workflows, addressing enterprise compliance concerns.

6. Admin Controls

Enterprise and Education administrators receive role-based controls to determine which users can create agents and which tools they can access. This allows organizations to deploy automation while maintaining security boundaries.

How This Differs from Custom GPTs

Custom GPTs, launched in November 2023, let users configure a version of ChatGPT with specific instructions and limited tool access. They were fundamentally reactive — responding to individual prompts in individual sessions.

Workspace Agents are proactive, persistent, and collaborative by design. They do not wait to be asked; they execute defined workflows on their own, remember what they have done, and improve based on team feedback. The shift is from a configurable chatbot to a deployable agent infrastructure.

OpenAI also confirmed that existing Custom GPTs can be converted into Workspace Agents, providing a migration path for organizations that have already built GPT-based workflows.

Use Cases

OpenAI's documentation highlights several representative use cases:

• Sales teams: Agents that prepare meeting briefs, pull CRM data, and draft follow-up emails automatically
• Engineering teams: Agents that monitor codebases, respond to GitHub review comments, and generate weekly summaries
• Operations: Scheduling agents that coordinate across calendars and send status reports
• Research: Agents that continuously scan specified sources and surface relevant updates

Pricing and Availability

Workspace Agents are currently free during the research preview period, which ends May 6, 2026. After that date, OpenAI will introduce credit-based billing. Specific per-credit pricing has not yet been disclosed, but the feature is restricted to paid business plans — free-tier ChatGPT users are not included.

Computer use capabilities within agents are currently unavailable in the EU and UK, reflecting ongoing regulatory alignment.

Competitive Context

This launch comes as the broader AI industry converges on agentic workflows as the next major product frontier. Anthropic's Claude is available through the API with tool use and multi-agent support. Google has been building agent infrastructure through Vertex AI. Microsoft's Copilot deeply integrates with Office 365 workflows.

OpenAI's differentiation is scale and distribution. ChatGPT has the largest consumer and business user base of any AI product, and embedding agents natively into that platform — especially with Slack integration — provides immediate reach that competitors would need to build from scratch.

Outlook

The credit-based pricing model that begins May 6 will be the first real market test of whether enterprise teams are willing to pay for persistent agent execution at scale. If the research preview builds strong adoption, OpenAI is positioned to turn ChatGPT into an enterprise workflow platform that competes not just with other AI tools but with RPA (Robotic Process Automation) software like UiPath and Automation Anywhere.

The key risk is reliability. Autonomous agents that act on behalf of teams carry a higher cost of failure than a chatbot giving a wrong answer. How well OpenAI's approval workflows and admin controls handle edge cases will determine whether this moves from research preview to mainstream enterprise deployment.

Conclusion

OpenAI's Workspace Agents mark a genuine step-change in what ChatGPT is: no longer a question-answering tool but an autonomous team member that runs in the cloud, integrates with existing tools, and handles end-to-end workflows. For teams already using ChatGPT Business or Enterprise, the research preview is worth exploring immediately. The May 6 pricing deadline gives organizations a limited window to assess value before committing to usage-based billing.