OpenAI Introduces Workspace Agents in ChatGPT for Team Workflows

OpenAI has added workspace agents to ChatGPT, a new feature aimed at teams that want shared AI workers with access to company tools and approvals. The company said the agents are available in research preview for ChatGPT Business, Enterprise, and Edu customers, and they are powered by Codex, OpenAI’s software-focused model and agent system.

The idea is to move beyond one-off prompts and into persistent assistants that can operate on behalf of a group. Instead of each employee creating their own private workflow, a workspace agent can be configured once and shared across an organization, with controls that determine what it can see, what it can do, and when it needs human approval.

OpenAI says these agents can work across ChatGPT and Slack, which makes them useful for common internal tasks that usually bounce between chat, documents, and software tools. In practice, that means a team could set up an agent to help triage requests, draft responses, pull together project context, or carry out structured workflows that would otherwise require a person to switch between systems.

The company is positioning workspace agents as a more governed way to use AI in business settings. That matters because many teams already use ChatGPT-style assistants informally, but those setups can become messy fast when different people create different prompts, connect different tools, and rely on inconsistent instructions.

A shared workspace agent gives administrators a way to keep those workflows inside company boundaries. OpenAI said the feature includes organizational controls and approvals, which suggests the agent does not get free rein over data or actions. For security and IT teams, that kind of gating is often the difference between a useful internal tool and another shadow-IT problem.

The research preview is also a signal that OpenAI still considers this an early product. Features in research preview usually mean the company wants real-world feedback while the system is still being tested, especially on edge cases around reliability, permissions, and how agents behave when they have access to multiple tools at once.

OpenAI is also using a pricing model that may shape adoption. The workspace agents are free until May 6, after which the company said they will move to credit-based pricing. That gives organizations a short window to experiment before usage becomes metered.

The broader push here is clear enough. OpenAI is trying to make ChatGPT less like a standalone chat box and more like an operating layer for team work, where agents can remember context, follow organizational rules, and coordinate across the systems people already use.

That fits a larger shift in enterprise AI, where buyers are no longer just asking whether a model can answer questions. They are asking whether it can sit inside existing workflows, respect permissions, and handle recurring tasks without turning into a governance headache. OpenAI’s answer, at least for this preview, is workspace agents built around shared control, approvals, and memory.

The feature is now available through ChatGPT for qualifying business and education customers, with Slack integration and managed access as part of the launch.