OpenAI turns ChatGPT into a real team workspace

Most people still use ChatGPT at work like a single-player game. You spend twenty minutes dialing in a prompt, copy the output into Slack, and the moment a coworker wants to build on it, they have to feed the exact same background context back to the AI from scratch.

OpenAI is trying to break that isolation. On September 25, the company rolled out shared projects and automated tool connectors for workplace plans. Alongside the release, OpenAI revealed that ChatGPT now has over 5 million paying business users.

Bringing multiplayer mode to workspaces

Shared projects are rolling out immediately to Business, Enterprise, and Edu tiers, with access coming to Free, Go, Plus, and Pro plans down the line. Enterprise and Edu workspaces have the feature switched off by default for admin control.

Instead of keeping instructions and docs locked to one user, team members can now dump reference materials and workspace instructions into a single shared project space. ChatGPT keeps track of that shared context, so every new conversation in the workspace automatically starts with the team's latest updates.

Projects also get their own isolated memory. If an account manager adds meeting notes, an engineer can jump into the project and immediately ask ChatGPT for a recap to prep a demo. Users can chat asynchronously or branch off a colleague's conversation to chase down a new idea, with permissions split between basic chat access and full edit rights.

Smarter connectors and enterprise guardrails

OpenAI also upgraded ChatGPT's connectors. The chatbot can now connect to Gmail, Google Calendar, Google Contacts, Microsoft Outlook, and Teams, alongside existing integrations like SharePoint, GitHub, Dropbox, and Box.

Crucially, ChatGPT can now automatically decide when to pull from an external connector, run a web search, or rely on internal knowledge. You no longer have to manually toggle integrations per chat. For GitHub, SharePoint, and cloud storage providers, data can be synced in advance to speed up responses.

To keep enterprise IT departments calm, OpenAI announced new ISO 27001, 27017, 27018, and 27701 certifications, an expanded SOC 2 report, role-based access control, OIDC single sign-on support, and optional IP allowlisting to block logins from outside company networks.

Why it matters

OpenAI is steadily moving from a standalone chatbot to a centralized enterprise work hub, stepping directly onto Microsoft 365 Copilot's home turf.

For distributed teams and small companies, shared projects eliminate the tedious chore of repeatedly pasting background docs and prompt guidelines across multiple seats. It is an immediate win for agency client work, unified content drafting, and internal reporting without needing a dedicated operations engineer to stitch tools together.

We think making connectors trigger automatically is the smartest part of this rollout. Toggling integrations manually was always too high of a friction point for casual enterprise users.

However, expanding deep into calendars, inboxes, and file systems creates inevitable security anxiety. Even with permission inheritance and strict data isolation, granting an AI engine access to company-wide communication channels will face pushback from conservative IT teams. Defaulting these features to disabled in Enterprise tiers shows OpenAI knows trust will take longer to build than features.