How to Use This Checklist

This checklist distills the pre-deployment preparation, technical setup, training program, and launch verification steps that we execute across every enterprise Claude deployment. It represents the lessons from 200+ implementations — including the hard-won discoveries about what gets skipped and what the skipping costs.

Not every item applies to every organization. Small companies (under 100 employees) can skip some of the governance infrastructure items. Organizations using the API directly rather than Claude.ai will need different technical checks. Use this as a starting framework and adapt to your context.

The checklist is organized into five phases matching our 90-day deployment methodology. Items marked CRITICAL are the ones we've seen cause deployment failures when skipped — don't skip them.

Want expert guidance through your deployment? Our implementation team has executed 200+ enterprise Claude deployments. Book a free 30-minute readiness assessment.
Request Free Assessment →

Phase 1: Pre-Deployment Assessment (Days 1-15)

This phase establishes the foundation. Skipping it and jumping straight to technical setup is the #1 cause of failed deployments — you end up with access configured but no one knowing how to use it safely or effectively.

Technical Assessment 10 items
CRITICAL: Inventory all existing AI tool usage (Claude, ChatGPT, Copilot, etc.) Employees often have personal accounts you don't know about. Discover them before you set policy, not after.
Document current network egress controls and determine if claude.ai/api.anthropic.com is accessible from corporate network
Assess SSO infrastructure (SAML 2.0 or OIDC availability) if deploying Claude.ai Teams/Enterprise
Determine API vs. claude.ai deployment model for each use case
Estimate API volume needs (requests/day, context window size) for rate limit planning
Identify MCP integration candidates (CRM, ticketing, code repos) and assess technical feasibility
Map compliance requirements (SOC 2, HIPAA, GDPR) to deployment decisions
CRITICAL: Determine if a Business Associate Agreement (BAA) is required for PHI handling
Assess data residency requirements and map to Anthropic's infrastructure
Document current productivity baseline metrics for ROI measurement post-deployment
Organizational Assessment 6 items
CRITICAL: Identify executive sponsor (VP or above) who owns the deployment and can unblock decisions
Map stakeholder support and resistance: who will champion, who will resist, who is neutral
Identify pilot department based on motivation, use case clarity, and feedback willingness
Assess existing AI literacy and Claude familiarity across target user population
Identify Claude Champion candidates in pilot department (use identification criteria from Champions Network guide)
Document pilot success criteria and measurement approach before beginning
📋
Free White Paper: Enterprise Claude Implementation Playbook Full deployment methodology, technical architecture patterns, and organizational change management frameworks.
Download Free →

Phase 2: Governance & Policy Setup (Days 10-25)

Governance Infrastructure 9 items
CRITICAL: Draft and review Claude-specific Acceptable Use Policy (not just general AI policy)
CRITICAL: Create data classification framework with Claude-specific guidance for each tier (Green/Yellow/Orange/Red)
Assign Claude Governance Owner (VP-level), Department Claude Leads, and Claude Security Contact
Define output verification requirements by output type and risk level
Build MCP Integration Approval workflow into existing request management system
Create incident response playbook for common Claude governance incidents
Execute Data Processing Agreement with Anthropic (required if processing EU personal data)
Execute Business Associate Agreement with Anthropic (required if processing PHI)
Schedule quarterly governance review calendar for the next 12 months

Phase 3: Technical Setup (Days 20-40)

Technical Configuration 11 items
Execute Anthropic enterprise agreement and configure billing/usage controls
Configure SSO integration and provision user access for pilot cohort
CRITICAL: Verify training data opt-out is confirmed in writing for all enterprise accounts
Configure zero-retention API mode for workflows involving sensitive data
Set up audit logging for Claude API calls in regulated-data workflows
Configure rate limits and usage alerts to prevent unexpected cost spikes
Build and test priority MCP integrations in staging environment
Create department Claude Projects with approved context documents and system prompts
Develop department-specific prompt template libraries (minimum 10 prompts per department)
Test pilot workflows end-to-end with sample data before user rollout
Verify network access from all user environments (office, VPN, remote)

Phase 4: Training & Launch (Days 35-60)

Training Program 7 items
CRITICAL: Complete Champion training before any general employee training begins
Deliver foundation training to pilot department with role-specific exercises
Launch pilot Slack/Teams channel for Claude questions and use case sharing
Run Day 3, Day 7, and Day 14 guided practice check-ins
Collect Day 30 adoption metrics and self-reported time savings survey
Document pilot learnings and update training curriculum before broader rollout
Prepare department-specific deep-dive training sessions for post-pilot rollout

Phase 5: Enterprise Rollout & Measurement (Days 60-90)

Enterprise Rollout 4 items
Provision access and deliver training for remaining departments in prioritized sequence
Establish cross-departmental Champions network with monthly meeting cadence
Communicate all-employee launch with governance policy, resources, and Champion contacts
Schedule Day 90 ROI assessment: adoption rates, time savings, quality outcomes

For a comprehensive deployment methodology with detailed guidance on each phase, see the Enterprise Claude Implementation Playbook or our Implementation Service page for guided deployment support.

Common deployment risks to monitor: low adoption in specific departments (usually indicates inadequate Champion coverage or training gaps), governance incidents (usually indicate data classification training gaps), and technical performance issues (usually indicate rate limit configuration or network routing issues). See our Change Management guide for handling adoption resistance.