ChatGPT for Teams — Business, Enterprise, seats and admin

Business seats

• Seats are capacity, not people. Owners buy Standard and/or Premium seats and assign them. You can move members between seat types and mix them freely.
• Minimum 2, maximum 200 paid seats (both types count).
• Beyond the included limits: buy workspace credits under flexible pricing (see OpenAI's ChatGPT rate card).
• OpenAI won't train on a Business workspace's data.

What Enterprise adds

From the pricing page: a larger context window, SCIM, Enterprise Key Management (EKM), user analytics, domain verification, role-based access controls, custom data-retention policies, data residency in ten regions, 24/7 priority support with SLAs, custom legal terms, invoicing, and volume discounts. Pricing is custom (credit-based or token-based). Enterprise also controls the planned custom-GPT retirement for its workspace; approved deferrals run to Feb 11, 2027.

Set up a Business workspace

1. Open the ChatGPT Business sign-up page, sign in with the account that should own the workspace, choose Business, and pick seat quantity, seat types and billing frequency (OpenAI). Buying extra Standard seats doesn't upgrade your own seat to Premium.
2. You become the workspace owner. Invite teammates by role and seat type. Add capacity later in Workspace settings → Billing and assign seats in Workspace settings → Members.
3. Decide whether to keep your personal workspace separate. Joining Business doesn't migrate your content or cancel a personal subscription.
4. Set ChatGPT Work access per role in Workspace settings → Permissions & roles (Work Cloud, Work Local, Codex Local), and starting models for Work & Codex in Workspace settings → Models.
5. Review memory settings and role permissions for your compliance needs.
6. Plan for custom GPTs retiring on Dec 11, 2026. Business workspaces can still create GPTs, but new workflows should go into plugins.

Admin habits that prevent the usual problems

• Deprovision promptly. The classic workspace failure is a leaver who still has access months later. Use SSO (Business supports SAML SSO and MFA) and remove members on day one.
• Mix seats deliberately. Put heavy Work and Codex users on Premium and everyone else on Standard, then review usage analytics monthly.
• Limit connected apps and plugins to what teams need. Each one is a path for workspace data to leave.
• Write a one-page internal guide: which Projects and plugins to use, the prompt patterns that work, and how to flag a problem.

Cost example: a 25-person team

• All Standard, billed annually: 25 × $20 = $500 a month ($625 billed monthly).
• 20 Standard + 5 Premium (annual): $400 + $500 = $900 a month.
• For comparison, 25 individual Plus plans cost the same $500 a month but come with no admin, no shared billing, and training on by default unless each person opts out.

When a ChatGPT workspace isn't the answer

• Data must stay on your own hardware: even Enterprise runs in OpenAI's cloud. Look at OpenClaw with local models or NemoClaw.
• You want several model vendors: ChatGPT is OpenAI-only. See Claude Cowork or Kilo Code.
• You want always-on agents under your control: ChatGPT Work covers scheduled and triggered tasks, but for self-hosted automation see Hermes.