Claude Enterprise rollout checklist

Choose the plan and seats

Moving from another assistant? See how to migrate from ChatGPT or Copilot to Claude.

Claude Team covers 2 to 150 people. Claude Enterprise adds the controls most companies need at scale. Both are priced per seat with annual billing, and Enterprise adds usage at API rates.

• Confirm the planTeam for 2 to 150 people. Enterprise when you need SCIM, audit logs, role-based permissions, custom retention, IP allowlisting or the Compliance API.
• Size the seats and the budgetEnterprise costs US$20 per seat per month, billed annually, plus usage at API rates. The self-serve minimum is 20 seats; plans bought through Anthropic’s sales team start at 50.
• Plan for usage costsUsage scales with the model and the task. Estimate it by role during the pilot, before rolling out to everyone.
• Know what a seat includesClaude on web, desktop and mobile, plus Claude Code and Cowork.
• Decide on HIPAAIf you handle protected health information, a HIPAA-ready plan is available. The Primary Owner accepts the business associate agreement, which cannot be undone, and Claude Code and Cowork need extra configuration.

Set up identity and access

• Verify and claim your domainsEnterprise can capture every account on your domains. Team can verify domains only.
• Turn on single sign-onWith SAML or OIDC, through your identity provider.
• Choose how accounts are createdSCIM keeps users and groups in sync with your identity provider, on Enterprise. Just-in-time provisioning creates an account at first sign-in.
• Define roles and permissionsRole-based permissions on Enterprise decide who can use which features.
• Restrict network access if requiredIP allowlisting on Enterprise limits access to your company networks.

Set data and security controls

• Set data retentionEnterprise supports custom retention periods, with a minimum of 30 days.
• Record how training worksBy default, Anthropic does not train its models on inputs and outputs from its commercial products, unless you send feedback or opt in. State this in your AI policy.
• Turn on audit logsAnd give your compliance team access to the Compliance API if they need programmatic access for monitoring.
• Decide on encryption keysEligible organisations can manage their own keys on AWS, Google Cloud or Azure.
• Check where data is processed and storedFor claude.ai, Anthropic stores data in the US. US-only inference is available on usage-based Enterprise plans at 1.1 times API rates. Workloads that must be processed in the EU can use Claude through Amazon Bedrock or Google Cloud EU endpoints.
• Complete a data protection impact assessment where requiredDocument the setup, the data involved and the controls you chose.

Connect tools, knowledge and skills

• Turn on approved connectorsSuch as Google Drive, Gmail, Google Calendar, GitHub, Microsoft 365 and Slack. An Owner has to enable them.
• Add custom connectors for internal systemsThrough remote MCP servers, each reviewed before it is turned on.
• Set up shared projectsProjects can be shared across the organisation, with the reference material each team needs.
• Publish organisation-wide skillsSkills package your company’s ways of working. They require code execution to be enabled.
• Decide on Claude Code and CoworkBoth come with the seat. Cowork is on by default, except in HIPAA organisations, and its desktop app is in beta.

Run a pilot

• Pick one or two departmentsWith leaders who will use Claude themselves.
• Choose three to five real workflows in eachAnd measure today’s time and quality as a baseline.
• Track adoption from the startUsage analytics are available on Team and Enterprise, with an Analytics API on Enterprise.
• Collect feedback every weekAnd adjust skills, connectors and instructions as you learn.

Train people and set the policy

• Write a short AI policyWhich data may be used, which tools are approved and how outputs are checked before they are used.
• Train by role, on real workUse each department’s own workflows, not generic demos.
• Cover AI literacy under the EU AI ActArticle 4 has applied since 2 February 2025: companies take measures to support the AI literacy of their staff. No certificate is required.
• Name champions in each departmentPeople who help colleagues and collect requests for new skills.

Measure adoption and scale

• Review at 30, 60 and 90 daysActive users, the workflows in use and time saved against the pilot baselines.
• Expand department by departmentUsing what the pilot taught you about skills, connectors and training.
• Review costs by teamSeats plus usage, against the value measured.
• Keep it currentNew skills and connectors, usage reports and policy updates every month.

Sources

• Claude Help Center: what is the Enterprise plan?
• Claude pricing
• Claude Help Center: set up single sign-on
• Claude Help Center: set up JIT or SCIM provisioning
• Claude Help Center: role-based permissions on Enterprise plans
• Claude Help Center: custom data retention controls
• Claude Help Center: audit logs
• Claude Help Center: customer-managed encryption keys
• Anthropic Privacy Center: is my data used for model training?
• Claude Help Center: connectors
• Claude Help Center: skills for your organisation
• Claude Help Center: HIPAA-ready Enterprise plans
• European Commission: AI literacy questions and answers