OpenAI Now Plans to Introduce AI Agents Into The Workplace

OpenAI is deliberately expanding its AI platform ChatGPT for professional use. With the newly introduced “Workspace Agents,” the company is gradually replacing the previous Custom GPTs with more powerful, continuously running agents designed to automate complex workflows within enterprises. The feature is aimed exclusively at business customers on Business, Enterprise, Edu, or Teachers plans.

As reported, OpenAI is undergoing a broader pivot away from end consumers toward the B2B space with enterprise customers — the very field where competitor Anthropic is currently making strong inroads with Claude Code and Claude Cowork.

The Transition from Custom GPTs to Agents

Custom GPTs, which users were previously able to configure themselves, are apparently regarded by OpenAI as a transitional technology. The company explicitly describes Workspace Agents as an “evolution of GPTs” and announces plans to actively facilitate the switch: existing GPTs will in future be convertible directly into Workspace Agents.

The key difference lies in the technical foundation and scope of use. While Custom GPTs primarily responded to individual conversational requests, Workspace Agents run continuously in the cloud, can execute tasks autonomously, and are designed for shared use within teams.

“Workspace Agents are an evolution of GPTs. Powered by Codex, they can take on many of the tasks that people already do at work, from generating reports and writing code to responding to messages.”

What Workspace Agents Can Actually Do

The agents are based on OpenAI’s Codex model and possess a range of capabilities that go well beyond simple chat functions. They can independently work through multi-step tasks, remember previous projects, and be connected to external applications.

• Scheduled execution: Agents can run on a schedule — for example, automatically generating reports every Friday and sending them to the team.
• App integration: Connections to external tools such as CRM systems, IT ticketing systems, and communication platforms are possible.
• Slack integration: Agents can be deployed directly in Slack channels to answer requests or receive tasks.
• Writing and executing code: Agents can not only write code but also execute it directly in a cloud environment.
• Memory across projects: Agents store context and learn from corrections over time.
• Approval workflows: For sensitive actions — such as sending emails or editing spreadsheets — agents can be configured to first obtain human approval.

Concrete Use Cases from Everyday Business Operations

OpenAI cites several examples already in use internally, intended to serve as a template for other companies:

• Software review: An agent evaluates software requests from employees, checks them against internal policies, and automatically creates IT tickets.
• Product feedback routing: Feedback from Slack, support channels, and public forums is collected, prioritized, and converted into weekly product summaries.
• Monthly accounting close: An agent handles parts of the monthly close, from journal entries to variance analysis, and produces audit-ready documentation.
• Sales lead qualification: Incoming leads are researched, scored, provided with personalized follow-up emails, and updated in the CRM.
• Supplier risk management: Agents screen third-party vendors for sanctions, financial risks, and reputational risks, and generate structured reports.

Control and Security for Enterprises

OpenAI emphasizes that companies retain full control over their agents. Administrators can define which data and tools an agent is permitted to use, who may create and share agents, and which actions require manual approval. In addition, there are built-in safeguards against so-called prompt injection attacks, in which external content attempts to manipulate the agent.

For Enterprise and Edu customers, a compliance API is also available, giving administrators complete transparency over the configuration, changes, and executions of all agents. Agents can also be deactivated by the administrator if needed.

Availability and Pricing Model

Workspace Agents are currently in a “Research Preview” and are available free of charge to users on Business, Enterprise, Edu, and Teachers plans. Starting May 6, 2026, OpenAI will switch to a credit-based pricing model. New agents can be created directly via the ChatGPT sidebar by describing a desired workflow in natural language.

“AI has already helped people work faster on their own. But many of the most important workflows within an organization depend on shared context, handoffs, and decisions across teams.”

With Workspace Agents, OpenAI is positioning itself squarely in competition with other technology giants such as Google, Microsoft, and Amazon, all of which are likewise investing billions in autonomous AI systems for enterprise use. The replacement of Custom GPTs marks a strategic shift: away from the individual chatbot toolkit, toward continuously running, team-capable automation tools.