OpenAI Launches Workspace Agents in ChatGPT for Enterprise Teams

OpenAI rolled out Workspace Agents in ChatGPT this week, replacing the company's earlier Custom GPTs framework with a shared, persistent, multi-tool successor designed for enterprise teams. The launch lands as the most direct play yet to turn ChatGPT into the operating layer for everyday business work.

Workspace Agents are available now in research preview to ChatGPT Business, Enterprise, Edu, and Teachers customers. The product is free through May 6, 2026, after which a credit-based pricing model takes over.

Why this matters

The first generation of ChatGPT for work was a chatbot. The second generation — Custom GPTs — gave teams a way to package prompts and small tool integrations. Both were fundamentally personal: each user invoked them, each session was isolated, and the work didn't persist between conversations.

Workspace Agents is a different category. The agents are shared, run in the cloud whether or not anyone is online, plug into the systems where work actually happens, and improve over time as teams refine them. That's a meaningful step closer to what enterprises have been asking for since 2023: AI that does the work, not AI that explains how to do it.

It also raises the strategic stakes for Microsoft Copilot, Google Workspace AI, and the standalone agent platforms from Salesforce, ServiceNow, and Adobe. OpenAI is now competing directly in the enterprise agent layer that those vendors have been building for the last two years.

What was actually announced

A Workspace Agent is a configurable AI worker that lives inside ChatGPT and can be invoked by any authorized member of an organization. Out of the box, agents can be connected to Slack, Google Drive, Microsoft 365, Salesforce, Notion, Atlassian Rovo, and a growing list of enterprise applications.

OpenAI shipped a template gallery to lower the configuration burden. Teams can pick from pre-built agents — sales meeting prep, weekly reporting, customer onboarding, ticket triage — and customize them for their environment, or build agents from scratch using a visual configuration flow.

The cloud execution model is a notable design choice. Once configured, an agent can run on a schedule or be triggered by events in connected systems. A sales agent can monitor a Salesforce pipeline overnight and email the team a synthesized morning briefing without anyone opening ChatGPT.

Agents can also be invoked from Slack directly. A user can mention an agent in a channel, ask it to pull data from a connected source, and receive the result inline — closing one of the longest-standing complaints about ChatGPT for work, which is that the work happens somewhere other than where the team is already collaborating.

Industry impact

For Microsoft, the launch is a competitive shot across the bow. Copilot's defining feature has been deep integration with Microsoft 365 — Word, Excel, Outlook, Teams. Workspace Agents now ship with first-class Microsoft 365 integration alongside equally capable Slack and Google Drive support, neutralizing one of Copilot's pricing-power advantages.

For Salesforce, the picture is more complex. Agentforce has been Salesforce's headline strategic bet. OpenAI's deep Salesforce integration in Workspace Agents means many customers will be able to get agentic capabilities on Salesforce data without buying an additional Salesforce SKU — putting pressure on Agentforce's pricing.

For the broader agent platform market, OpenAI is essentially saying that an enterprise can run its agent layer inside ChatGPT, with credit-based pricing and a familiar interface. That's a meaningfully different positioning from Adobe's CX Enterprise, Microsoft's Copilot Studio, or Google's Gemini Enterprise, all of which require deeper platform commitments.

Expert perspectives

VentureBeat called Workspace Agents the successor to Custom GPTs and emphasized that the integration depth — Slack, Salesforce, Notion, Microsoft 365 — is what differentiates this release from the prior generation of ChatGPT enterprise features. The New Stack framed the launch as ending "the friction of manual team handoffs," noting the always-on cloud execution model is closer to a workflow engine than a chatbot.

Gartner published a quick first-take observing that Workspace Agents introduce new governance hurdles for enterprise IT — particularly around credential delegation, audit logging, and how agents handle data that crosses regulatory boundaries. That governance frame is the most important constraint on enterprise rollout speed and will determine which customers can adopt aggressively.

The skeptical read is that "research preview" still means research preview. Enterprises that have been burned by overpromising AI agent demos will want to wait for production maturity before committing to workflows that the agents are running unattended.

What's next

Watch the May 6 pricing transition. OpenAI's choice of credit-based pricing — rather than per-seat — is a signal that the company expects agent workloads to scale super-linearly with team size. Customers will need to model usage carefully before the meter starts running.

Watch the integrations roadmap. The current connector list is strong but missing some categories — particularly customer support platforms beyond Salesforce, and developer tooling like GitHub and Linear. Expect OpenAI to fill those gaps quickly given the competitive pressure from Microsoft and Google.

Watch the governance maturation. Enterprise IT and security teams will be looking for clearer answers on how agents authenticate to connected systems, how their actions are audited, and whether their behavior can be constrained at the policy level. OpenAI's enterprise momentum depends on those answers being convincing.

The bottom line

Workspace Agents is OpenAI's most serious enterprise product to date. It's not just ChatGPT plus integrations — it's a real attempt to make ChatGPT the place where work runs, not just where it gets discussed. For teams already using ChatGPT Business, Enterprise, or Edu, this is worth a serious look during the free preview window. For everyone else, the launch is a clear signal that the agent platform competition just got tighter.