On April 22, 2026, OpenAI launched something that quietly shifted what ChatGPT is. Not an update to the chat interface or a new model release. A fundamentally different product built for teams.
ChatGPT Workspace Agents are shared AI agents that businesses can build once, deploy across their organization, and let run automatically in the background, handling complex workflows while employees focus on other things. They connect to Slack, Google Drive, Salesforce, Microsoft 365, Notion, and dozens of other tools teams already use every day.
If you are on a business plan and wondering what Workspace Agents actually are, how they differ from what came before, and whether they are worth paying attention to, here is the full breakdown.
What Are ChatGPT Workspace Agents?
Workspace Agents are persistent, cloud-based AI agents that operate inside your organization’s ChatGPT environment. Unlike a regular ChatGPT conversation that exists only while you are using it, Workspace Agents run continuously in the background. They can keep working even when you close your laptop or step away from your desk.
They are powered by Codex, OpenAI’s coding and reasoning model, which gives them the ability to write and run code, connect to external tools, and execute multi-step workflows that go far beyond answering questions.
More Read
Here is what makes them different from everything OpenAI has built before:
They can be shared across an entire team. One agent built by a manager or administrator becomes available to everyone in the workspace. Improvements made to the agent benefit the whole team immediately.
They take actions inside the tools your business already runs. Not just inside ChatGPT. They can post to Slack channels, read and update Salesforce records, pull data from Google Drive, draft and send emails through Gmail, and update spreadsheets in Microsoft 365, all without a human manually triggering each step.
More Read
They run on schedules or triggers. You can set an agent to produce a weekly report every Monday morning, route incoming support tickets every hour, or fire automatically when a new lead appears in your CRM.
They remember context across sessions. Unlike a standard ChatGPT conversation that starts fresh each time, Workspace Agents maintain memory of what they have learned and done, getting more useful as teams build on them over time.
How Workspace Agents Differ From Custom GPTs
If you have been using ChatGPT for a while, you likely know about Custom GPTs, the personalized AI assistants OpenAI launched in late 2023 that let users create scoped versions of ChatGPT for specific roles or workflows.
Workspace Agents are not an incremental improvement on Custom GPTs. They are a different category of tool.
Custom GPTs are essentially configured chat assistants. You give them a persona, a set of instructions, and optionally a few custom actions. They respond when you prompt them. They wait. They do not act on their own.
Workspace Agents are automated workers. They do not wait for you to start a conversation. They can initiate tasks based on a schedule, a trigger from a connected tool, or a workflow condition you set. They write and execute code. They pull live data from connected systems. They take action inside external apps and report back.
More Read
The simplest way to put it: Custom GPTs are assistants that answer questions. Workspace Agents are systems that do the work.
OpenAI has signaled that Custom GPTs are effectively entering maintenance mode for business accounts. A one-click conversion tool from Custom GPT to Workspace Agent is coming, but no date has been confirmed yet. Individual users who built Custom GPTs on personal accounts can continue using them indefinitely. Business, Enterprise, Edu, and Teachers accounts will eventually be required to migrate.
What Tools Do Workspace Agents Connect To?
At launch, Workspace Agents ship with integrations for the most widely used business platforms:
Communication and collaboration: Slack is the most heavily promoted integration. An agent can live inside a company Slack channel, respond to mentions, follow threads, and take ownership of tickets without the team needing to leave Slack.
Google Workspace: Gmail, Google Drive, Google Calendar, Google Docs, and Google Sheets are all supported, allowing agents to read, write, and update content across the full Google productivity suite.
Salesforce: Agents can read CRM records, update deal stages, pull account data, and route information through sales workflows.
Microsoft 365: Word, Excel, and other Microsoft apps are included in the integration set.
Notion: Documentation, project tracking, and knowledge base management are supported through the Notion connection.
Atlassian Rovo: For teams using Jira and Confluence, the Atlassian integration enables agents to create and update tickets and pull project context.
OpenAI has said the connector list will expand over time. The platform also supports custom MCP servers for teams that need to connect agents to proprietary internal systems.
Real Use Cases Already in Production
OpenAI has published specific examples of Workspace Agents that its own internal teams built and are actively using. These give the clearest picture of what the technology looks like in practice.
Software Request Reviewer. When an employee submits a software access request, this agent reads the request, checks it against the company’s approved tools and security policies, recommends the appropriate next steps, and automatically files an IT ticket. What previously required a human reviewer to handle manually now runs without manual input.
Product Feedback Router. This agent monitors Slack channels, customer support tickets, and public review forums for product feedback. It organizes that feedback into prioritized tickets and delivers weekly summaries to the product team. It runs continuously, not when someone remembers to check.
Weekly Metrics Reporter. Every Friday, this agent pulls data from connected business systems, creates charts, writes a narrative summary of the key numbers, and shares the finished report with the relevant team. The entire reporting workflow happens automatically.
Lead Outreach Agent. When a new inbound lead appears in the CRM, this agent researches the lead, scores them against the company’s qualification criteria, drafts a personalized follow-up email, and updates the CRM record. Sales representatives at OpenAI report this saves five to six hours per week in account research and outreach preparation.
Third-Party Risk Assessment Agent. For compliance workflows that require reviewing vendor or partner information against risk criteria, this agent handles the initial screening and flags cases that need human review.
How to Build a Workspace Agent
OpenAI designed the creation process to be accessible without technical skills. To get started:
Click Agents in the ChatGPT sidebar. Describe the workflow you want to automate in plain language, or drop in a file that describes the process. ChatGPT guides you through the remaining steps: defining the sequence of actions, connecting the relevant tools, testing the agent until it works as expected, and publishing it to your workspace for team use.
Pre-built templates are available for common use cases across finance, sales, marketing, and operations. These come with built-in tool connections and suggested steps, so teams can customize from a working starting point rather than building from scratch.
Once an agent is live, it appears in the shared workspace where any team member with the appropriate permissions can use it. Improvements made to a shared agent are available to everyone using it immediately. Because agents can be guided and corrected through conversation, they improve over time as teams interact with them and refine their behavior.
Who Can Use Workspace Agents and What Does It Cost?
Workspace Agents are available on the following plans:
- ChatGPT Business at $20 to $25 per user per month depending on billing terms
- Enterprise at custom pricing
- Edu and Teachers plans
They are not available on Free, Go, Plus, or Pro individual plans. This is a deliberate product decision. Workspace Agents are designed for organizational use with admin controls, role-based permissions, and shared team infrastructure. Individual plans do not have the team management layer these agents require.
On pricing: Workspace Agents were free to build and use during the initial research preview period that ran through May 6, 2026. From that date, OpenAI moved to a credit-based pricing model where each agent action consumes credits from the organization’s monthly pool. The exact per-credit cost has not been publicly disclosed, with pricing varying based on task complexity, the number of tools accessed, and execution time.
What Is Coming Next
OpenAI has outlined several additions to Workspace Agents that are in development:
New triggers that can start agent work automatically based on events in connected tools, not just schedules set in advance. Better admin dashboards to give organizations visibility into how agents are performing, where they are being used, and how to optimize them. More integrations with business tools beyond the launch set. Support for Workspace Agents inside the Codex app, which handles coding and technical workflows.
The conversion path from Custom GPTs to Workspace Agents is also coming. OpenAI has promised a one-click migration tool that will let teams convert their existing GPT library into Workspace Agents without having to rebuild from scratch.
Why This Launch Matters Beyond OpenAI
The April 22, 2026 launch did not happen in isolation. On the same day, Google announced the Gemini Enterprise Agent Platform, and Salesforce expanded its partnership with Google Cloud to connect its Agentforce product with Gemini. Three major technology companies launched competing enterprise agent products on the same day.
That is not a coincidence. It reflects where the enterprise AI market is heading in 2026. The question for business technology is no longer whether companies will use AI. It is which platform will become the operating system for the automated workflows that run inside organizations.
OpenAI’s bet with Workspace Agents is that ChatGPT, already embedded in millions of workplaces as an individual productivity tool, can extend that footprint into the team automation layer by making it easy to build agents that connect to the tools teams already use.
The Bottom Line
ChatGPT Workspace Agents represent the most significant product evolution in ChatGPT’s history for business users. They move the platform from a chat tool that individuals use in isolation to a team automation system that can handle entire workflows across an organization’s existing software stack.
For teams on Business or Enterprise plans that have repeatable, time-consuming workflows, especially anything involving regular reporting, lead processing, ticket routing, or cross-tool data management, Workspace Agents are worth building and testing now while the platform is still in its early stages and teams that get familiar with it early will have a meaningful advantage.
