OpenAI launches ChatGPT workspace agents for team workflows

The company is positioning agents as a team layer above individual AI assistance

OpenAI is expanding its push from personal productivity into coordinated workplace automation with the launch of workspace agents in ChatGPT, a new product the company says is designed for shared, long-running tasks inside organizations.

Announced April 22, the feature is being introduced as a research preview for ChatGPT Business, Enterprise, Edu, and Teachers plans. OpenAI describes workspace agents as an evolution of GPTs: Codex-powered agents that can be created once, shared across a team, and used to handle multi-step workflows such as preparing reports, drafting messages, writing code, routing requests, or moving work across connected systems.

The core claim is not simply that AI can help an individual user produce text faster. It is that teams can package a recurring workflow into an agent that works within organizational permissions, pulls context from the right tools, asks for approval when required, and continues operating in the cloud even when the user is offline.

From solo prompting to shared process automation

That distinction is important. Most mainstream generative AI adoption to date has been framed around the individual knowledge worker: summarize this document, draft this email, write this code snippet. Workspace agents target a different level of value. They are meant to sit closer to business process infrastructure, where the challenge is not one person’s output, but coordination among people, systems, approvals, and handoffs.

OpenAI’s own framing emphasizes exactly that point. The company says many of the most important workflows inside organizations depend on shared context and cross-team decisions, and that workspace agents are designed for those situations. It cites an internal use case in which its sales team uses an agent to gather details from call notes and account research, qualify leads, and draft follow-up emails directly in a representative’s inbox.

If that model works broadly, it could mark a meaningful step in enterprise AI adoption. The value proposition shifts from “AI as assistant” to “AI as workflow participant,” one that can gather information, follow predefined steps, and keep tasks moving rather than waiting for a human to reassemble context each time.

AI & Robotics

OpenAI is opening ChatGPT for Clinicians at no cost to verified U.S. physicians, nurse practitioners, physician assistants, and pharmacists, pairing the launch with a new benchmark for clinical chat tasks.

DT Editorial AI·Apr 25, 2026·via openai.com

AI & Robotics

OpenAI is offering up to $25,000 for a universal jailbreak that defeats a five-question biology safety challenge in GPT-5.5, turning external red teaming into a focused test of frontier-model safeguards.

DT Editorial AI·Apr 25, 2026·via openai.com

AI & Robotics

Cohere’s planned acquisition of Aleph Alpha is more than a startup takeover. It is a bid to build a politically backed sovereign AI supplier for governments and regulated industries in Europe and beyond.

DT Editorial AI·Apr 25, 2026·via the-decoder.com

AI & Robotics

OpenAI says GPT-5.5 is built to handle longer, tool-using workflows with less supervision, extending the company’s push toward agent-style AI systems for coding, research, and data work.

The competitive enterprise question: permissions and control

OpenAI is also clearly aiming at the governance question that has slowed some enterprise adoption. The company says workspace agents operate within the permissions and controls set by the organization. That framing matters because companies are often less worried about whether an AI model can draft a report than whether it can do so safely, using approved systems, with clear boundaries around access and approvals.

The ability to share an agent across a workspace is another important piece. Consumer AI tools often scale poorly inside organizations because each user recreates prompts, workflows, and conventions independently. A shared agent gives teams a reusable object: one workflow definition, many users, and the possibility of improving it over time. OpenAI also says teams will be able to use these agents in ChatGPT or Slack, suggesting the company wants agents to live where work already happens rather than only inside a standalone interface.

Editor’s note in the announcement said GPTs will remain available while teams test workspace agents, and that OpenAI plans to make it easy to convert GPTs into workspace agents. That signals an evolutionary product path rather than an abrupt replacement. Existing custom AI setups are being positioned as building blocks for more organizationally aware automation.