ChatGPT for Teams & Business — Workspace Setup, Roles & Cost

Tier comparison

Pricing varies by region, annual commitment, and seat count. Get an actual quote — OpenAI is generous with multi-year discounts at the Enterprise tier.

Setup walkthrough

1. Create the workspace

For Business: go to chatgpt.com/business, click Get Business, complete checkout. The account you use becomes the initial Owner.

For Enterprise: contact OpenAI sales — Enterprise involves a custom DPA, security questionnaire, and (typically) a 1-month onboarding window. Worth the wait for any regulated industry.

2. Invite members

1. Settings → Members → Invite.
2. Add member emails (one per line). Bulk paste works.
3. Assign role at invite time (default: Member). Skip this if you'll use SSO/SCIM for provisioning.
4. Send. Members get an email with a one-click join link.

3. Configure SSO (Business and Enterprise)

SSO is strongly recommended — it lets you de-provision a leaving employee in seconds rather than having to remember to remove them manually.

1. Settings → Workspace → SSO.
2. Pick your provider (Okta, Google Workspace, Microsoft Entra/Azure AD).
3. Follow the SAML/OIDC config — ChatGPT shows you the exact ACS URL and Entity ID to paste into your IdP.
4. Test with one user before enforcing org-wide.
5. (Enterprise only) Set up SCIM for automatic provisioning/deprovisioning.

4. Publish a shared Custom GPT

1. Build the Custom GPT as usual (see Custom GPTs deep dive).
2. In the GPT settings, set Sharing to Anyone in [Workspace name].
3. Save. The GPT now appears in every workspace member's sidebar under "Workspace GPTs."
4. (Recommended) Add the GPT to Featured via Workspace settings so it shows at the top of everyone's GPT list.

5. Configure admin policies

• Data controls: Settings → Workspace → Data controls. Business and Enterprise are excluded from training by default — verify the toggle is off.
• Memory: Disable Memory org-wide if compliance prohibits cross-conversation context. Settings → Workspace → Memory.
• External GPT access: Settings → Workspace → External. Decide whether members can use public Custom GPTs from outside the workspace, or only workspace-published ones.
• Retention: (Enterprise) Settings → Workspace → Retention. Set how long conversation history is stored. Default is "indefinite"; many compliance regimes require shorter (90 days, 1 year).
• Audit logs: (Business basic, Enterprise full) Settings → Workspace → Audit logs. Configure SIEM forwarding if you have one.

Common admin patterns

• One admin per 50 seats. Below that, one admin is enough. Above, distribute the work so no one becomes a single point of failure.
• "Featured GPTs" should be 3–5, not 30. Featuring everything is featuring nothing. Pick the workflows every member should use weekly; let the rest live in the regular workspace GPT list.
• Onboarding doc > "go play with it." A 1-page internal doc that says "here are the 3 GPTs you'll use, here's the prompt pattern that works, here's how to flag a problem" drives 10× the adoption of just rolling it out.
• Off-boarding via SSO, not manual. The #1 ChatGPT-for-business security failure mode is forgetting to remove someone who left 4 months ago. SSO with auto-deprovisioning eliminates this.
• Audit the public Custom GPT toggle. If members can use public GPTs, they may be feeding workspace data to third-party developers. Either disable, or have a clear policy.

Cost worked example — 25-seat team

Hypothetical mid-size team on Business at $30/seat/month:

• 25 seats × $30 = $750/mo base cost
• Heavy users may push the tier's per-tool-call quota — budget 10–20% overage on top in months when teams do major research or migration work
• Compared to giving everyone a personal Plus subscription: 25 × $23 = $575/mo. Business adds ~$175/mo for the admin controls, SSO, and shared GPTs. Worth it for anything >5 seats.
• Compared to Enterprise: typically 30–60% more than Business per seat, but unlocks SCIM, custom retention, and DPA terms that compliance teams will actually approve.

See the cost calculator for an interactive comparison against per-API-token pricing if you're considering building on the OpenAI API instead.

When Business / Enterprise isn't the right choice

• You have <5 seats and no compliance need: individual Plus subscriptions work fine. Save the admin overhead.
• You need full data residency and zero-egress guarantees: even Enterprise data lives in OpenAI's cloud. If that's a deal-breaker, look at self-hosted alternatives like OpenClaw + Ollama or NemoClaw on your own GPU.
• You want model flexibility: ChatGPT is OpenAI-only. Teams using Claude, Gemini, or mix-and-match should consider Claude Cowork or Kilo Code.
• You need long-running async agents: ChatGPT is conversational. For nightly tasks, scheduled workflows, and multi-day projects, see Hermes or OpenClaw.