How to Roll Out Claude to Your Whole Team: A Setup Playbook

Rolling Claude out to a team is a different problem from using it yourself. The product is the same, but suddenly there are seats to provision, single sign-on to wire up, connectors to approve, Skills to push organization-wide, and pitfalls that don’t exist on Pro or Max. The good news: Anthropic now publishes a clear admin path, and the work is mostly checklist-driven. This guide walks through what each plan gets you, the order admins should set things up in, and the traps that catch most teams during their first rollout.

Everything below is sourced from Anthropic’s own admin documentation at support.claude.com, claude.com/pricing, and the Cowork enterprise blog. Pricing and feature claims are current as of May 2026.

Team vs Enterprise: Pick the Right Plan First

Anthropic offers two plans designed for organizations, and they’re not interchangeable. Per claude.com/pricing:

Team plan

• Standard seats: $20/seat/month annual ($25 monthly). Each standard seat gets 1.25× the per-session usage of Pro.
• Premium seats: $100/seat/month annual ($125 monthly). 5× the usage of standard seats — about 6.25× Pro per session — and Claude Code is bundled.
• Seat range: 5 minimum, 150 maximum. Above 150 you’re on Enterprise.
• Usage limits are per-member, not pooled. One person’s heavy usage doesn’t drain anyone else’s allowance.

Team gets the things Pro and Max can’t have: central billing, SSO, JIT provisioning, role-based permissioning, organization spend caps, enterprise search across connectors (Google Drive, Gmail, GitHub, Microsoft 365, Slack), admin controls for connectors, the enterprise deployment of the Claude desktop app, and — critically — no model training on your content by default.

Enterprise plan

Per claude.com/pricing/enterprise: $20/seat plus separate usage charges, billed annually. Two flavors:

• Self-serve: usage is purchased upfront in credits. When credits run out, service stops until topped up. Seats can’t be removed mid-term.
• Sales-assisted: usage billed monthly in arrears. Custom MSA, volume discounts implied. Mid-term seat changes negotiable.

Enterprise adds on top of Team: SCIM directory sync, audit logs, the Compliance API for real-time programmatic access to usage data, fine-grained role-based access, custom data retention controls, IP allowlisting, network access control, HIPAA-ready configuration, and Claude Security (in beta).

Rule of thumb: under 150 users without a strict compliance program, Team is sufficient. Above 150, or with hard SCIM and audit-log requirements, Enterprise is mandatory. For broader plan context our Claude plans guide walks through the consumer tiers too.

The Admin Console — Where Owners Live

The Team and Enterprise admin console lives at claude.ai/admin-settings/organization. It’s where Owners and Primary Owners manage members, billing, Skills, connectors, SSO, and provisioning. Per Anthropic’s SSO setup article, the canonical sections are:

• Organization settings → Members — add members directly, in bulk, or by SCIM auto-provisioning. Assign role and seat type (Standard or Premium).
• Organization settings → Connectors — admin-enable connectors before users can attach them.
• Organization settings → Skills — toggle Skills as a capability for the org and push organization-wide Skills.
• Authentication — set up or manage SSO.
• User provisioning — set up or manage SCIM (Enterprise only).
• Spend caps — at organization level and individual user level.
• Usage analytics — Cowork sessions, active users, DAU/WAU/MAU metrics.

Anthropic’s recommended setup order, per the SSO setup wizard: review prerequisites → verify your domain(s) → set up SSO with your IdP → choose whether to require SSO → choose your provisioning approach (JIT or SCIM).

SSO and User Provisioning

Per Anthropic’s user-provisioning docs, supported identity providers include Okta SAML, Microsoft Entra ID SAML (formerly Azure AD), Google SAML, OneLogin SAML, JumpCloud SAML, Duo SAML, plus additional IdPs through WorkOS (Anthropic’s SSO infrastructure provider).

Two provisioning approaches, with very different operational implications:

• JIT (Just-In-Time) — available on Team and Enterprise. Members assigned to the Anthropic app in your IdP get an account created automatically the first time they log in. Lightweight; works on Team.
• SCIM directory sync — Enterprise only. Members are auto-provisioned by directory sync without needing first login. Required if you want true headcount-aligned licensing.

Two things to flag before enabling SSO enforcement. First, per Anthropic’s SSO enforcement warning: once “Require SSO for Claude” is on, “users not added to SSO will be unable to access their existing Free/Pro/Team/Max accounts” on that email. The accounts aren’t deleted, but they’re inaccessible until those users are added to SSO. Tell employees in advance to export anything they want to keep.

Second, with SCIM: “Saving before users are properly assigned will result in those users being deprovisioned from the organization.” Configure your IdP user assignments before flipping the SCIM toggle, not after.

Org-Level Connector Controls

Connectors are how Claude reaches into the apps your team already uses — Google Drive, Gmail, GitHub, Slack, Microsoft 365, and so on. On Team and Enterprise, the install path is two-layer: an admin enables a connector at the org level first, then individual members authenticate with their own accounts via OAuth. Our broader Skills, Connectors, and Plugins guide goes deeper on the conceptual side.

Microsoft 365 deserves special attention because it has a unique third step. Per Anthropic’s M365 connector docs, “A Microsoft Entra Global Administrator in your tenant must authorize the connector” before any organization Owner can enable it on the Claude side. Block out time for that handoff — your Microsoft admin and your Claude admin are likely two different people, and the IT ticket can take longer than the Claude-side configuration.

Admins can also restrict which actions are allowed within a connector at the org level — letting Claude search Gmail but not send messages, or read Drive but not modify files. This goes beyond per-user OAuth scopes; it’s an org-wide policy gate.

Org-Wide Skills and Cowork Controls

Per Anthropic’s Skills admin docs, organization-wide Skills are deployed under Organization settings → Skills → Organization skills → +Add. You upload a .zip containing a SKILL.md file, and “the skill is immediately provisioned to all users in your organization.”

Skills come in three layers: personal (uploaded by individual members), shared with you (peer-shared), and organization (admin-provisioned, enabled by default for everyone, optional toggle off per user). Important caveat: admins can push org Skills, but they cannot centrally manage personal Skills uploaded by individual users. Each member still owns their personal Skills library.

Cowork went generally available on all paid plans on May 5, 2026, and the GA brought a set of org-level controls per claude.com/product/cowork:

• Role-based access controls — group users (manually or via SCIM), assign each group a custom role defining which Claude capabilities they can use.
• Group spend limits — per-team budgets configured in the admin console with predictable cost ceilings.
• Usage analytics — Cowork session counts, skill and connector invocations, and DAU/WAU/MAU metrics in the admin dashboard plus an Analytics API.
• OpenTelemetry observability — pipe Claude usage data into your existing observability stack.
• Per-connector action restrictions — read-only vs. write at the org level.

Privacy and Compliance

Anthropic states the policy plainly on its enterprise product page: “Customer data on Claude Enterprise is not used to train our models.” The Privacy Center reinforces this for Team: “if you’re using Claude for Work [Team], Claude Enterprise, Claude for Education, or Claude Gov, your data is not used for model training — that’s covered under Anthropic’s commercial terms and is not subject to opt-in or opt-out toggles.”

Compliance certifications, per privacy.claude.com (last updated March 16, 2026): HIPAA-ready (BAA available), ISO 27001:2022, ISO/IEC 42001:2023 (the AI-specific management standard), and SOC 2 Type I and Type II. Data residency options and Zero Data Retention mode are documented in Anthropic’s Trust Portal at trust.anthropic.com — talk to your account team if either is a deal-breaker.

Anthropic’s Own Rollout Playbook

Anthropic published a deployment guide titled “Deploying Claude across the enterprise with Claude Cowork”. The structure they recommend:

• A six-month roadmap for moving from champion teams to organization-wide deployment.
• A five-level Cowork maturity model ranging from chat-and-Q&A to department-wide plugins.
• Starting with early adopters in finance, legal, sales, and product.
• Choosing your first use case carefully and structuring a pilot evaluation before scaling.
• An “integration-first” principle: Claude Cowork “meets work where it already lives — local files and folders, connected apps like Slack and Google Drive, and the browser.”

Customer references in Anthropic’s own materials include Thomson Reuters, Zapier, Jamf, and PayPal (with a published webinar walking through the PayPal deployment). For regulated industries, the financial-services deployment guide PDF on Anthropic’s CDN is a good starting point.

Pitfalls That Catch First-Time Admins

• Mid-cycle seat removal doesn’t refund. On Team, removing a member doesn’t refund the seat — the seat is reusable until renewal. Plan headcount changes around the renewal cycle.
• SSO enforcement locks out non-SSO users from any prior Free/Pro/Team/Max account on that email. Communicate the change at least a week before flipping the toggle.
• SCIM is Enterprise only. If you’re on Team and need true directory sync, you’ll need to upgrade.
• SCIM saves can deprovision users if you save before assigning them in your IdP. Configure assignments first.
• Microsoft 365 connector requires three approvals in sequence: Microsoft Entra Global Admin → Claude Owner → end user OAuth. Each is a different person at most companies.
• Custom Skills are user-owned, not admin-managed. Admins can push org-level Skills but can’t reach into a user’s personal Skills library.
• Enterprise self-serve seats can’t be removed mid-term. Only sales-assisted contracts allow that flexibility — pick the right contract type at the outset.
• Owner/Primary Owner roles are powerful. Build the custom-role groups (on Enterprise) before turning members loose, so capabilities are scoped from day one.

A Pragmatic Setup Order

1. Pick the plan: Team if under 150 users and no SCIM/audit requirements, Enterprise otherwise.
2. Provision the org. Verify domain(s) so colleagues can find your org during signup.
3. Wire up SSO with your IdP. Don’t enforce it yet — verify a few test logins first.
4. Add a Microsoft Entra ticket if you’ll use the Microsoft 365 connector. That handoff is usually the longest-tail step.
5. Enable Skills as a capability and push your first org-wide Skills (a brand template, a writing voice guide, a financial-modeling skill).
6. Approve connectors at the org level. Pick the 2-3 your team actually uses daily; ignore the rest until people ask.
7. Run a 2-week pilot with one champion team. Track usage in the admin analytics dashboard.
8. Communicate the SSO enforcement date with at least a week of lead time. Then enforce.
9. Open access to the broader org. Enable spend caps before you do.
10. Review usage analytics monthly. Move power users to Premium seats; reclaim seats from non-users at renewal.

You May Also Like

• Claude Max vs Pro vs Free: Which Plan Is Worth It?
• Skills vs Connectors vs Plugins: How to Extend Claude
• Claude Usage Limits: How They Work and How to Stretch Them Further
• Claude Projects: Organize Your AI Workspace for Maximum Productivity
• How to Use Claude AI: The Complete Beginner’s Guide