OpenAI has launched workspace agents in ChatGPT, reducing the need for organizations to build custom agent systems from scratch and making AI agents configurable and deployable at the workflow level, allowing business teams, operations leaders, and non-technical users to adopt AI without relying on dedicated engineering resources. Available now in research preview for ChatGPT Business, Enterprise, Edu, and Teachers plans, workspace agents are shared, cloud-based agents that business teams, operations leaders, and non-technical users can build, assign, and run across real workflows without writing code or standing up custom infrastructure.

For most organizations, deploying AI agents has meant a development project first and a business tool second. Workspace agents change that equation by moving AI task delegation from something teams have to engineer into something they can configure and deploy directly within business workflows. Agents are free to eligible plan holders until May 6, 2026, when credit-based pricing begins.

In short, workspace agents are shared, cloud-based AI systems that execute multi-step tasks across business workflows, allowing teams to automate work without building custom infrastructure.

A workspace agent operates as a cloud-based AI system that executes multi-step workflows on behalf of a team, working across connected tools and data sources while staying within administrator-defined permissions.

Workspace agents are shared, cloud-based AI systems that execute multi-step workflows across business tools, allowing teams to automate work without building custom infrastructure or relying on engineering resources.

Businesses have been hearing about AI agents for months, but actually deploying them has typically required engineering resources, custom development, and ongoing technical maintenance. OpenAI is changing that equation. Workspace agents in ChatGPT are shared, cloud-based agents that business teams can build, assign, and run across real workflows without writing code or standing up infrastructure, all while operating within the permissions and controls set by their organization.

AI has already helped individuals work faster on their own, but many of the most important workflows inside an organization depend on shared context, handoffs, and decisions across teams. Workspace agents are built for exactly that kind of work. They can gather context from the right systems, follow team processes, ask for approval when needed, and keep work moving across tools, without requiring manual coordination at each step. OpenAI's own sales team uses an agent to pull together details from call notes and account research, qualify new leads, and draft follow-up emails directly in a rep's inbox, letting account teams spend less time stitching together information and more time with customers.

Workspace agents evolved from GPTs, the custom AI tools OpenAI introduced previously, and are powered by Codex, giving them the ability to take on tasks people already do at work, from preparing reports to writing code to responding to messages. They run in the cloud, so they can keep working even when no one is actively present. They are also designed to be shared across an organization, so teams can build an agent once, use it together in ChatGPT or Slack, and improve it over time. Existing GPTs will remain available while teams evaluate workspace agents, and OpenAI says conversion tools are coming.

Workspace agents are functional and available to eligible plan holders today, but the research preview designation means the product will continue to evolve based on real-world use before becoming available more broadly.

At the technical core of workspace agents is Codex, OpenAI's code-capable AI system, which gives agents the ability to do more than respond to prompts. Agents powered by Codex can write and execute code, interact with connected applications, retain memory across sessions, and continue working across multiple steps without requiring a user to stay present.

That persistent operation is what makes workspace agents practically useful beyond a single task. Agents can be set to run on a schedule — pulling a weekly report every Friday, for example — or deployed inside Slack channels where they pick up incoming requests, respond with answers and linked documentation, and file tickets when they identify unresolved issues. Teams stay unblocked faster, and critical follow-ups are less likely to fall through the gaps.

Each agent has access to a cloud-based workspace that includes files, code, connected tools, and memory, giving it everything it needs to execute multi-step workflows, retain context between sessions, and take action across connected systems.

When sensitive actions are involved, agents can be configured to request human approval before proceeding, keeping people in the decision loop while routine work continues autonomously.

In most organizations, process knowledge lives across people, documents, and institutional memory, and walks out the door when someone leaves. Workspace agents give teams a way to capture that knowledge as a reusable workflow, one that follows established processes, draws on the right tools, and can be shared across the entire organization.

OpenAI's own accounting team built an agent that handles key parts of month-end close, including journal entries, balance sheet reconciliations, and variance analysis. The agent completes the work in minutes, generates workpapers with the underlying inputs and control totals needed for review, and follows internal policies throughout.

Because agents have memory and can be corrected through conversation, they improve as teams use them, making it practical to build once and adapt for new workflows over time.

The key point: workspace agents are not chat assistants being repurposed for tasks. They are purpose-built workflow executors that hold context, use tools, and operate continuously; the difference being that they now require configuration rather than custom code to deploy.

Building an agent starts with a plain-language description of the job to be done, or by dropping in a file. A team member describes the workflow they want automated — qualifying inbound leads, routing customer feedback, preparing a weekly metrics report — and ChatGPT guides the setup from there, defining the steps, connecting the right tools, adding skills, and testing the agent until it performs as expected.

OpenAI provided several examples of agents already running inside the company's own teams, illustrating the range of functions workspace agents are designed to handle.

These examples span sales, IT, product, and finance, reflecting OpenAI's intent to make workspace agents applicable across business functions, not just technical teams.

Templates are also available across finance, sales, marketing, and other functions, giving teams a faster starting point with pre-built skills and suggested tool connections. Training and documentation are available through OpenAI Academy and the OpenAI Help Center.

Workspace agents include enterprise-grade monitoring and controls built in, giving admins the tools to protect sensitive data without slowing teams down. ChatGPT Enterprise and Edu admins can determine which connected tools and actions specific user groups can access, manage who has permission to build, use, and share agents, and monitor all agent activity through the Compliance API. The Compliance API provides visibility into every agent's configuration, updates, and run history. Agents can also be suspended if needed.

For sensitive actions — editing a spreadsheet, sending an email, or adding a calendar event — agents can be configured to request human approval before proceeding, keeping people in the decision loop for consequential steps while allowing agents to handle routine work autonomously.

Built-in safeguards address prompt injection, a class of attack in which malicious instructions embedded in external content attempt to redirect an AI model's behavior, helping agents stay aligned with their original instructions even when processing content from outside the organization.

Usage analytics are also available after an agent is shared, giving teams visibility into how many runs have completed and how many people are actively using each agent.

Looking ahead, admins will also be able to view every agent built across their organization directly in the admin console, including usage patterns and connected data sources, giving organizations a fuller picture of how AI is being deployed across teams.

OpenAI shared results from several early testers, and two accounts in particular illustrate what workspace agents can deliver in practice.

At Rippling, AI Engineering lead Ankur Bhatt described how a single Sales Consultant built, evaluated, and iterated a full agent without any engineering support:

"The hard part of building an agent is not the model. It's the integrations, memory, the user experience. Workspace agents collapsed that work, so one of our Sales Consultants built, evaluated, and iterated a Sales Opportunity agent end to end without an engineering team. It researches accounts, summarizes Gong calls, and posts deal briefs directly into the team's Slack room. What used to take reps 5-6 hours a week now runs automatically in the background on every deal." — Ankur Bhatt, AI Engineering, Rippling

At Hibob, Senior Director of AI and Innovation Alon Arbiv described a similar outcome for executive scheduling:

"One of our teams built a workspace agent for executive scheduling, refining it with real user feedback and no engineering support. The agent now handles conflicts, incoming requests, and calendar flow across channels, so leaders stay focused without getting pulled into the details. What used to be a daily juggling act now runs quietly in the background, giving executives more space to focus on bigger priorities." — Alon Arbiv, Senior Director AI & Innovation, Hibob

Both accounts point to the same outcome: meaningful time savings on routine, high-context tasks, built and deployed by the people closest to the work, freeing technical teams to focus on higher-order work.

Workspace agents are available now in research preview for ChatGPT Business, Enterprise, Edu, and Teachers plans. Admins on Enterprise and Edu plans can manage access through role-based controls, determining which users can build, use, and share agents within their organization.

Pricing is straightforward for now: workspace agents are free to eligible plan holders until May 6, 2026, when credit-based pricing takes effect. OpenAI has not yet detailed the specific credit structure.

The current release is also just the starting point. OpenAI says additional capabilities are coming in the weeks ahead, including new triggers that can start agent work automatically, improved dashboards for tracking and refining agent performance, expanded options for agents to take action across a wider range of business tools, and support for workspace agents inside the Codex app.

Q: What are OpenAI workspace agents and how are they different from GPTs?
A: In short, workspace agents are AI systems that complete ongoing work for teams, not just respond to prompts. They execute multi-step workflows on behalf of a team, operating continuously in the background or on a schedule across connected tools. Unlike GPTs, which handle individual interactions, workspace agents are designed to retain context and keep work moving without constant user input. OpenAI says existing GPTs can be converted into workspace agents.

Q: Why do workspace agents matter for businesses?
A: Workspace agents reduce the need for organizations to build custom AI systems. Instead of requiring engineering resources and development time, teams can configure and deploy agents directly within their workflows, lowering the barrier to adoption and allowing business teams to use AI more immediately and operationally.

Q: How do workspace agents work inside an organization?
A: A team member describes the workflow they want the agent to handle, and ChatGPT guides setup by defining steps, connecting tools, and configuring capabilities. Once deployed, the agent can run in ChatGPT or Slack, operate on a schedule, execute multi-step tasks, and request human approval when needed. Codex powers the agent's ability to carry out complex workflows and maintain context across sessions.

Q: Where can workspace agents run today?
A: Workspace agents can currently run in ChatGPT and Slack, allowing them to operate directly inside the tools where teams already communicate and manage work. OpenAI has indicated that support for additional platforms is planned.

Q: Who can build and use workspace agents?
A: Any team member can create a workspace agent using natural language, with ChatGPT guiding the setup process. This makes it possible for non-technical users to build and deploy agents without writing code.

Q: How do organizations control and manage workspace agents?
A: Admins can control which tools and actions agents can access, manage permissions for different user groups, and monitor agent activity, including through the Compliance API. Agents operate within defined safeguards, including role-based controls and prompt injection protections, and can be suspended at any time.

Q: What are the current limitations of workspace agents?
A: Workspace agents are currently in research preview and available only on ChatGPT Business, Enterprise, Edu, and Teachers plans. Deployment is currently limited to ChatGPT and Slack, and credit-based pricing begins May 6, 2026, with full pricing details still forthcoming. Some administrative features, including centralized visibility across all agents, are still being developed.

Organizations have spent the past year evaluating AI agents as a concept while waiting for a practical entry point that does not require a development project first.

Key point: OpenAI's workspace agents do not resolve every open question about enterprise AI deployment, but they materially lower the threshold for getting started. Organizations can now move from identifying a workflow to running an agent on it without a development cycle in between.

Who should care: Business leaders and operations teams who have been waiting for AI agents to become accessible without engineering investment should pay close attention. This is directly relevant to any organization running repeatable, high-context workflows across sales, finance, HR, IT, or operations.

Why this matters now: The release arrives as organizations are actively evaluating where AI fits into their operating model. Workspace agents provide a concrete entry point — an operational tool available today for eligible plan holders — rather than something limited to pilots or proof-of-concept experiments.

What decision this affects: For organizations currently building or planning to build a custom agent system, workspace agents should be evaluated as a potential alternative. For those that have not yet started, this lowers the primary barrier that has kept agents at the engineering level.

In short, OpenAI is turning AI agents into operational tools that organizations can assign work to, not systems they need to build before they can use.

The practical question for business leaders is no longer whether AI agents are ready. It is whether their team's workflows are.

Editor’s Note: This article was created by Alicia Shapiro, CMO of AiNews.com, with writing support from Claude, and AEO/GEO/SEO optimization, image concept development, and editorial structuring support from ChatGPT, AI assistants. All final editorial decisions, perspectives, and publishing choices were made by Alicia Shapiro.