OpenAI has launched workspace agents inside ChatGPT, a significant upgrade from the Custom GPTs many teams have been using for the past two years. The new agents are powered by Codex, run persistently in the cloud, and are designed to be shared across an entire organisation rather than set up by individual users.

The feature is available now in ChatGPT Business, Enterprise, Edu, and Teachers plans as a research preview. It is free until May 6, 2026, after which credit-based pricing kicks in.

What Workspace Agents Actually Do

The core difference from Custom GPTs is autonomy and persistence. A Custom GPT waited for you to start a conversation. A workspace agent runs in the background, keeps going after you close the browser, and can operate across multiple tools at once.

Teams configure an agent once, then everyone in the organisation can use it through ChatGPT or directly from Slack. The agents connect to external tools including Salesforce, code repositories, and other work systems, letting them gather the right context, take action, and report back.

Use cases OpenAI highlighted include preparing weekly reports, responding to routine messages, writing and reviewing code, and handling multi-step research tasks that would otherwise require multiple manual hand-offs between people.

The agents are also designed with approval checkpoints built in. If an action is sensitive or outside the agent’s configured permissions, it pauses and asks a human before proceeding. That guardrail matters for any business thinking seriously about where AI fits into operations versus where it needs oversight.

Why This Is a Step Change From Custom GPTs

Custom GPTs were essentially a packaging layer on top of ChatGPT. You could give them instructions and knowledge, but they were still essentially chatbots waiting for prompts. They also lived on your individual account and could not be centrally managed or improved by a team.

Workspace agents change that in three ways. First, they are persistent, meaning they do not reset between sessions. Second, they are organisational, meaning IT admins and team leads can govern them, update them, and see what they are doing. Third, they are connected, meaning they can actually touch your business systems rather than just talking about them.

This aligns with the broader direction across the industry. The dominant pattern in 2025 was building clever prompts. In 2026, the shift is to agents with tools, memory, and real-world actions.

Pricing Lands on May 6

The free window closes on May 6. After that, workspace agents consume credits from your ChatGPT plan. OpenAI has not published a detailed credit pricing table yet, but the implication is that usage-intensive agents, such as those running daily report generation or continuous monitoring tasks, will carry meaningful ongoing costs.

For finance teams evaluating AI spend, this is worth modelling before the free period ends. Heavy usage during the preview phase should give a reasonable proxy for what monthly credit consumption will look like in production.

What This Means for Business

The launch of workspace agents confirms something that has been obvious for a while: OpenAI is not just a model company. It is building enterprise workflow infrastructure that competes directly with established players like Microsoft Copilot, Salesforce Agentforce, and Atlassian Rovo.

For business owners and operations leaders, the practical upshot is that the barrier to deploying task-specific AI automation inside your team just got lower. You no longer need a developer to wire up an agent to Slack or Salesforce. The integration layer is already there.

The harder question is not whether to use these tools but how to deploy them well. Teams that get the most value will be the ones who know how to identify the right workflows, write clean agent instructions, and build in the right oversight checkpoints. That is a judgment and design skill, not just a technical one.

Enterprise DNA works with businesses on exactly this: figuring out where AI agents create genuine leverage and where they create noise. If your team is thinking about adding AI agents to core workflows this quarter, a structured conversation about which processes are actually worth automating tends to save significant time and cost.

For a deeper walkthrough of tools like this and how they fit together, the free Working With Claude field guide covers the ecosystem end to end. Get the guide.