Before You Start: What You Need to Decide

Setting up Claude well requires three decisions before you touch any settings. Getting these right upfront saves significant rework later.

Decision 1: SaaS or API? If your team will be doing knowledge work — drafting, analysis, research, summarization — start with Claude.ai Team or Enterprise. No engineering required, and you can be live in days. If you need Claude embedded in custom applications or automated workflows, you'll need the API. Most enterprise teams start with SaaS and layer in API integrations later.

Decision 2: Which tier? Claude.ai Team works for departments of 5–40 that want shared Projects and consolidated billing. Claude.ai Enterprise is for larger deployments or organizations with compliance requirements (SSO, audit logs, custom retention). Contact Anthropic sales for Enterprise pricing — it's negotiated based on seat count.

Decision 3: Who is the admin? Identify one person who will manage the Claude deployment for your team — provisioning access, configuring Projects, being the go-to for policy questions. In smaller teams this is often the team lead. In larger organizations, IT or a designated Claude Champion takes this role.

Week One: Provisioning and Configuration

The technical setup for Claude.ai Team or Enterprise is straightforward. Here's the exact sequence we recommend.

1
Create Your Organization Account and Provision Seats
Sign up at claude.ai, navigate to Settings → Organization, and set up your org. For Team tier, add members by email. For Enterprise, work with your Anthropic account manager to configure SSO/SAML — this typically takes 1–3 days for IT to set up on your identity provider (Okta, Azure AD, Google Workspace). Don't add all users immediately — start with 5–10 early adopters for your initial configuration phase.
2
Configure Your First Claude Project
Create a Claude Project for your team's primary use case. In Project settings, write a system prompt that gives Claude context about your organization, the team's role, and the output standards you expect. Upload 2–3 reference documents that Claude will need: your style guide, key terminology document, or relevant policy document. Name the Project clearly (e.g., "Legal Contract Review" or "Marketing Content Assistant") so team members understand its purpose.
3
Draft Your Usage Policy (One Page Maximum)
Before you roll out to the full team, document your organization's acceptable use guidelines for Claude in one page or less. Cover: what data can go into Claude (and what can't), what outputs require human review before external use, how to handle confidential client/customer information, and who to contact with questions. This doesn't need to be a legal document — a clear, plain-language page is more useful and more likely to be read.
4
Test With Your Early Adopters (Days 3–5)
Give your 5–10 early adopters access and ask them to complete 5 real work tasks using the configured Project. Collect their feedback: What worked? What did Claude get wrong? What would make the Project configuration more useful? Use this feedback to refine your system prompt and any uploaded reference documents before full rollout. Early adopter testing consistently reveals prompt configuration gaps that would otherwise frustrate your broader team.
5
Full Team Rollout and Kick-Off Workshop
Once configuration is tested, provision remaining team members and run a 90-minute kick-off workshop. Structure: 20 minutes on what Claude is and your org's usage policy, 30 minutes of live demonstrations using your actual work (have 3 prepared examples), 30 minutes of hands-on practice where everyone completes one real task using Claude, 10 minutes for Q&A. Record the session for team members who miss it.

Want expert-led setup for your team? Our Training & Enablement service covers configuration, system prompt design, usage policy drafting, and a facilitated team kick-off workshop.

Request Free Assessment →

Weeks 2–4: Building the Habit

Technical setup is only half the challenge. The harder part is building the habit — getting your team to use Claude consistently for the tasks where it adds the most value. Here's what distinguishes high-adoption deployments from low-adoption ones.

The single most effective adoption driver is making Claude the path of least resistance for specific, high-frequency tasks. If your legal team writes 20 contract summaries a week, create a "Contract Summary" Project template that's pre-configured for exactly this task. Make it so easy to use Claude for this task that it's more effort not to use it than to use it. Abstract away the prompting skill requirement for the first use case — give your team a template they can use immediately without needing to understand prompt engineering.

Assign a Claude Champion — the person on your team who's most enthusiastic about Claude and willing to be the internal expert others go to. This person should be given explicit time (we recommend 2–3 hours per week in the first month) to experiment with Claude, build out the team's prompt library, and answer colleagues' questions. In our deployment experience, teams with an active Champion have 3–4x higher 90-day adoption rates than teams without one.

Run a 30-minute "wins sharing" session at the end of week two and week four. Ask team members to share one thing they used Claude for that saved them meaningful time. These sessions build social proof within your team — hearing that a respected colleague saved two hours on a task is more motivating than any training material. Capture the prompts used in the wins and add them to your team's growing prompt library.

Enterprise Claude implementation guide
Free Download

Enterprise Claude Implementation Playbook

Our complete 90-day deployment playbook includes team setup templates, system prompt examples for 10 departments, usage policy template, and adoption measurement framework.

Download Free →

Measuring Whether Your Setup Is Working

By the end of week four, you should be measuring three things. First, adoption rate: what percentage of your team is using Claude at least three times per week? For a healthy deployment, this number should be 50–60% at four weeks and climbing. If it's below 40%, something is blocking adoption — usually a gap in the initial training, a usability issue with your Project configuration, or a mismatch between the use cases you've focused on and the work your team actually does most.

Second, measure time saved on a representative sample of tasks. Pick five tasks your team does frequently, time them manually for one week before Claude and one week after. Average time savings of 25–40% on targeted tasks in the first month is typical for a well-configured deployment. If you're not seeing at least 20% savings on any task, revisit your prompt configuration — the issue is almost always that Claude's output requires too much revision, which points to a prompt design problem rather than a model capability problem.

Third, track the output revision rate: of the Claude outputs your team produces, what percentage require significant revision before use? Target less than 30% in month one. If it's higher, run a retrospective with your early adopters to understand the most common revision patterns — these patterns will tell you exactly how to improve your system prompt and task prompts.

30-Day Setup Checklist

  • Organization account created, seats provisioned (SSO configured for Enterprise)
  • At least one Claude Project created with a system prompt specific to your team's use case
  • Reference documents uploaded to Project (style guide, terminology, key policies)
  • One-page usage policy drafted and shared with team
  • Early adopter testing completed, system prompt refined based on feedback
  • Full team kick-off workshop run (recorded for absentees)
  • Claude Champion identified and given dedicated time to support the deployment
  • Initial prompt library created with 5–10 tested templates for your team's top use cases
  • Wins sharing session run at end of week 2 and week 4
  • Adoption and time-savings metrics measured at the 30-day mark

Five Setup Mistakes That Kill Adoption (And How to Avoid Them)

Mistake 1: No system prompt. Teams that use Claude without a configured system prompt get generic responses that don't reflect their organization's voice, standards, or context. Every team deployment should have at least a basic system prompt. Fifteen minutes of system prompt configuration saves hours of revision across thousands of uses.

Mistake 2: Too many use cases at launch. The teams that try to use Claude for everything from day one often end up using it for nothing well. Pick one or two high-frequency, high-value use cases for your first 30 days. Win clearly there, then expand. Focused deployment beats broad experimentation every time.

Mistake 3: No training on what not to put in Claude. Most teams understand that Claude is useful. Fewer understand the data governance boundaries. If someone puts sensitive customer data or regulated health information into Claude without appropriate controls in place, it creates a compliance issue that sets back the entire deployment. Cover data handling explicitly in your kick-off workshop.

Mistake 4: Skipping the Champion model. Leaving adoption to individual initiative produces mediocre results. Every successful deployment we've run has a Champion who actively drives usage, answers questions, and improves the prompt library. If you're not willing to give one person dedicated time to this role, your adoption will plateau early.

Mistake 5: Not measuring anything. Without measurement, you can't demonstrate ROI, can't identify what's not working, and can't build the business case for expanding the deployment. Set up your measurement framework before you roll out — it takes 30 minutes and makes everything else easier.