OpenAI launches workspace agents for ChatGPT

OpenAI has introduced workspace agents for ChatGPT, cloud-run assistants powered by the Codex model. The feature is available as a research preview for Business, Enterprise, Edu and Teacher ChatGPT subscriptions and is free until May 6, when a credit-based billing model begins.

Users create a workspace agent in ChatGPT by selecting Agents in the sidebar, describing the task or uploading a file, and following guided steps to define actions, connect tools and test the agent. OpenAI provides templates for finance, sales and marketing workflows, and shared agents update immediately so every team member sees improvements.

Workspace agents can access files, write or run code, use connected applications, store memory and perform multi-step workflows on a schedule or in response to requests. Because they run in the cloud, agents can continue tasks when individual users are offline. OpenAI wrote: “Powered by Codex, they can take on many of the tasks people already do at work — from preparing reports, to writing code, to responding to messages.”

OpenAI gave practical examples of use. One agent reviews software requests, checks them against tools and policies and files IT tickets. Another monitors Slack and public forums to collect product feedback and summarizes findings. A weekly report agent can pull data, generate charts and draft a report, and a lead outreach agent can research inbound leads, draft follow-up messages and update a CRM.

Administrators can set policies to limit which actions, tools and data an agent may access. For sensitive steps such as editing spreadsheets, sending emails or adding calendar events, organizations can require human approval. Usage is tracked through analytics and a compliance API provides administrators visibility into how agents are built and used, with the option to suspend agents if necessary.

Workspace agents integrate with ChatGPT and Slack at launch, and additional integration points are planned. OpenAI described the agents as an evolution of GPTs and said it will provide tools to convert existing GPTs into workspace agents.

The release follows other enterprise AI agent products introduced this year. OpenAI added that agents improve over time because they retain memory and can be guided and corrected in conversation: “Because agents have memory and can be guided and corrected in conversation, they get better as teams use them.”