OpenAI flipped a switch today that changes what ChatGPT actually is. The company launched Workspace Agents in research preview, and the short version is this: ChatGPT stopped being a chatbot and became something much closer to an autonomous employee.
This is not a minor update. It matters for small businesses in a very practical way. Here is what happened and what you should do about it.
What Workspace Agents Actually Do
Regular ChatGPT waits for you to ask it something, then answers. Workspace Agents do not wait. They run on a schedule, watch for triggers, pull data from connected apps, take actions across tools, and keep going even when you close the browser.
OpenAI is calling this "an evolution of GPTs." The key difference is that these agents are designed for teams, not individuals. Each agent has its own persistent workspace with file access, memory, connected tools, and the ability to write and run code. They can be hooked directly into Slack, where they pick up requests on their own.
The examples OpenAI shared publicly are telling. A "Lead Outreach Agent" researches incoming leads, scores them, drafts personalized follow-up emails, and updates your CRM automatically. A "Weekly Metrics Reporter" pulls data every Friday, builds charts, and sends a report to the team without anyone touching it. A "Product Feedback Router" watches Slack and support channels, converting feedback into prioritized tickets and weekly summaries.
For a small business, that is three full workflows you no longer have to run manually.
How You Set One Up
Building an agent is closer to describing a workflow than writing code. You describe a recurring process in plain language, or upload a file explaining it, and ChatGPT builds the agent: defining steps, connecting tools, adding skills, and testing the result. OpenAI also offers pre-built templates for finance, sales, and marketing to get you started faster.
Agents can connect to third-party apps including Slack, and more integrations are expected as the research preview matures. Existing custom GPTs are not being killed off; OpenAI says a conversion tool to upgrade them into full Workspace Agents is in progress.
The Pricing Window You Should Not Miss
Right now, Workspace Agents are free during research preview. Access is available on ChatGPT Business, Enterprise, Edu, and Teachers plans.
That changes on May 6, 2026, when OpenAI switches to credit-based pricing. The exact cost per credit has not been announced, but the pattern is consistent with their other API-adjacent products: the earlier you experiment and understand what you actually need, the less you pay when the meter starts running.
If you are already on a ChatGPT Business plan, you have roughly 12 days to build agents, test them against real workflows, and figure out which ones actually save you meaningful time. That is the only honest way to evaluate whether the cost-per-credit model will pencil out for your operation.
What This Means for Small Business Owners
The honest read: this is the most significant shift in what AI tools can do for small teams since the original GPT-4 launch. The gap between "AI assistant" and "automated workflow" just narrowed considerably.
The practical read: most small businesses are not running sophisticated enough tooling yet to take full advantage of multi-app agents on day one. But the lead outreach and reporting use cases are accessible right now without complex integrations. Those alone are worth the time investment this week.
The cautionary read: agents that act autonomously can also make mistakes autonomously. OpenAI built in approval checkpoints for high-stakes decisions, and you should use them. Start with agents that surface recommendations and ask for a human sign-off before they send emails or update records.
The free window is short. The upside is real. Build something this week.
Sources: The Decoder, TechStrong AI, IT Pro, OpenAI