OpenAI Just Turned ChatGPT Into a Team Automation Platform With Workspace Agents
ChatGPT isn’t just a chatbot anymore. As of yesterday, it’s a full-blown team automation platform — and the implications for how businesses use AI just shifted dramatically.
Workspace Agents
ChatGPT’s Evolution from Chatbot to Team Platform
Cloud-Based
Works while you’re
offline or asleep
Codex-Powered
Write code, run scripts,
execute multi-step tasks
Slack Integration
Deploy agents directly
in team channels
Scheduled Runs
Set agents to work on
recurring schedules
Team Sharing
Build once, share
across the org
Tools Stack AI | April 2026
OpenAI announced Workspace Agents on April 22, 2026 — a new feature that lets teams create shared, Codex-powered agents inside ChatGPT that can handle complex tasks, run in the cloud autonomously, and operate across tools like Slack. Think of it as Custom GPTs that grew up and got a job.
What Changed From Custom GPTs
If you’ve used Custom GPTs before, workspace agents will feel familiar at first glance. But the upgrade is significant. Custom GPTs were basically chatbot personas — you could give them instructions and they’d respond differently, but they still needed you sitting there typing prompts to do anything.
Workspace agents break that limitation entirely. They run in the cloud, which means they keep working even when you close your laptop and go home. They can execute code, connect to external applications, maintain memory across interactions, and handle multi-step workflows that span hours or even days.
The biggest shift? These agents are team-oriented. You build one, share it across your organization, and your entire team can use it. No more every employee creating their own janky Custom GPT for the same task.
How Workspace Agents Actually Work
Here’s what I found interesting about the architecture. Workspace agents are powered by Codex under the hood, giving them the ability to write and run code as part of their workflows. That means they’re not just answering questions — they can pull data from APIs, process spreadsheets, generate reports, and execute scripts.
The setup process is deliberately simple. You click “Agents” in the ChatGPT sidebar, describe a workflow your team does regularly, and ChatGPT walks you through building the agent step by step. Once it’s created, you can deploy it in two ways: inside ChatGPT itself or directly in Slack where your team already communicates.
Scheduling is baked in from day one. You can set agents to run on recurring schedules — daily reports, weekly data pulls, hourly monitoring checks. It’s the kind of functionality that would normally require tools like Zapier or n8n to set up, and OpenAI is building it directly into ChatGPT’s interface.
What You Can Build With Them
OpenAI shared several use cases that give a sense of the scope:
- Report preparation: An agent that pulls data from connected apps, processes it, and delivers a formatted report on a schedule
- Code review assistant: An agent deployed in Slack that reviews pull requests and provides feedback when tagged
- Customer message triage: An agent that monitors incoming messages, categorizes them, and drafts responses
- Documentation updater: An agent that watches for code changes and automatically updates related documentation
The common thread across all these: they’re not one-shot tasks. They’re ongoing workflows that traditionally required either a dedicated employee or a custom software integration to handle.
ChatGPT’s Evolution: From Chatbot to Automation Platform
Nov 2022
ChatGPT
Basic chat
Nov 2023
Custom GPTs
Personal bots
Feb 2025
Codex Launch
Cloud compute
Apr 2026
Workspace Agents
Team automation
Key Shift:
Custom GPTs = personal assistants that answer when asked
Workspace Agents = team workers that execute autonomously
Tools Stack AI | Source: OpenAI announcements
Who Gets Access (and What It Costs)
Workspace agents are launching as a research preview for ChatGPT Business, Enterprise, Edu, and Teachers plans. Consumer ChatGPT Plus and Free users don’t get this yet — it’s clearly positioned as a business product.
Here’s the pricing angle that matters: workspace agents are free to use until May 6, 2026. After that, OpenAI will switch to credit-based pricing. They haven’t published exact rates yet, but given how Codex pricing works elsewhere, expect costs to be tied to compute usage rather than flat fees.
Existing Custom GPTs can eventually be converted to workspace agents, though OpenAI says that conversion feature is coming “in the future” rather than at launch.
How This Compares to the Competition
This announcement puts OpenAI in direct competition with several categories of tools simultaneously. Workflow automation platforms like Zapier and n8n suddenly have to contend with agents that can do similar things but with natural language instructions instead of drag-and-drop builders. Team AI assistants like Lindy AI and Relevance AI lose some differentiation. And Anthropic’s Claude, which recently launched its own scheduled routines feature, now has to match the team-sharing and Slack deployment capabilities.
What’s different about OpenAI’s approach is distribution. ChatGPT already has over 900 million weekly active users and more than a million enterprise customers. Building workspace agents directly into the product that teams already use gives OpenAI a massive adoption advantage. You don’t need to buy a new tool or convince your IT department to approve another vendor — it’s just a new tab in ChatGPT.
What This Means for Your Team
If you’re running a team that does any kind of repetitive knowledge work, workspace agents are worth testing during the free preview period. The barrier to entry is essentially zero — you describe what you want automated and ChatGPT builds the agent for you.
The biggest practical impact will be for small and mid-size teams that can’t afford dedicated automation engineers. A three-person marketing team can now have an agent that monitors competitor pricing daily, generates weekly performance reports, and drafts social media responses — all running in the background while the team focuses on strategy.
For larger enterprises, the security controls and organizational permissions are the selling point. Workspace agents respect your existing ChatGPT Enterprise controls, which means IT can manage what agents have access to without opening up new security surface area.
My Quick Take
I’ve been saying for months that the real battle in AI isn’t about which model is smartest — it’s about which AI platform becomes the default operating system for work. OpenAI just made the biggest move yet in that direction. Workspace agents transform ChatGPT from a tool you open when you have a question into a platform that’s working for your team 24/7.
The free preview period is smart. It gets teams hooked before pricing kicks in. If the Codex-powered execution is reliable enough for production workflows (and based on what I’ve seen, it’s getting there), this could become one of those features that makes ChatGPT Business genuinely hard to cancel.
That said, I want to see how the credit pricing shakes out after May 6. If running a workspace agent on a schedule costs $50/month in compute credits, the value proposition is incredible. If it’s $500/month, it starts competing with dedicated automation tools that might be more capable for specific use cases.
FAQ
Are workspace agents available on ChatGPT Plus?
Not yet. Workspace agents are currently available as a research preview only on ChatGPT Business, Enterprise, Edu, and Teachers plans. There’s no timeline for when individual Plus subscribers might get access.
How much will workspace agents cost after the free period?
OpenAI will introduce credit-based pricing starting May 6, 2026. Exact rates haven’t been published yet, but agents will likely be billed based on compute usage (similar to Codex pricing), meaning costs will scale with how much work your agents do.
Can I convert my existing Custom GPTs to workspace agents?
OpenAI has confirmed this conversion feature is planned but it’s not available at launch. For now, you’ll need to recreate your workflows as workspace agents from scratch using the new agent builder in the ChatGPT sidebar.
