Skip to main content

Claude Training Plan

This plan turns the Adoption & Governance Guide into four concrete trainings plus the trainer guidance and checklists to deliver them consistently.

Didactic idea (all trainings): value first, limits and risks second, practical examples over product theory, clear takeaways at the end, concrete next steps for daily work. Rules of thumb: max 4–5 key points per slide, one practical example per topic block, never name a risk without an action, never demo a feature without its governance context.

Training matrix​

#TrainingAudienceDurationCore message
1Using Claude safely in the companyAll employees20–25 minA useful tool, never a replacement for responsibility
2Setting up Claude for teamsTeam leads, owners, multipliers25–30 minTeams need structure, not just access
3Claude Code & Cowork safelyIT, Dev, Security, Digital Office, power users25–30 minPowerful functions need stronger control
4Configuring Claude Enterprise as adminIT admins, IAM, Security, platform owners~30 minIdentity and roles before features; test before you enforce

A shared closing slide for trainings 1–3: Claude is a work tool; security and data protection always apply; standards make teams productive; agentic functions need special control; results stay a human responsibility.

Schulung 1 — Using Claude safely in the company​

Audience: all employees · Duration: 20–25 min · Goal: understand safe, productive basic use.

Core message: Claude is a useful work tool, but no substitute for responsibility. Only approved data, only through the company access, always check results.

Learning objectives — after the training, participants can:

  • name typical allowed uses,
  • recognize disallowed uses,
  • apply the data traffic light,
  • recite the five everyday rules.

Outline:

  1. Intro & value — drafts, summaries, routine work
  2. Rules & limits — allowed, not allowed, the traffic light
  3. Practical examples — Sales, Marketing, Support, Management
  4. Close & takeaways — 5 rules, check before use, contacts

Participant checklist:

[ ] I use Claude only through the company access
[ ] I never enter passwords, tokens, or secrets
[ ] I check the data class before entering anything
[ ] I use only approved extensions and sources
[ ] I check results before using or forwarding them
[ ] I ask when I am unsure

Closing question: Which one mistake will you avoid with Claude from tomorrow?

This training maps directly to the Onboarding page — hand it out as the follow-up.

Schulung 2 — Setting up Claude for teams​

Audience: team leads, business owners, multipliers · Duration: 25–30 min · Goal: set up team use in a structured way.

Core message: Teams need structure, not just access. Projects, Skills, and clear owners make Claude reproducible and scalable.

Learning objectives — participants can:

  • distinguish Projects, Skills, Connectors, Plugins, Artifacts,
  • define a small team pilot,
  • name governance roles for their team.

Outline:

  1. Why team structure matters — avoid sprawl, create standards
  2. The building blocks — Projects, Skills, Connectors, Plugins, Artifacts
  3. Team examples — Sales, IT, Marketing, Management, Support
  4. Governance & best practices — owners, reviews, versioning, start small

How a team should start: 1–2 clear use cases → one Project per topic → 1–3 Skills → only necessary Connectors → name an owner for each Project and Skill.

Participant checklist:

[ ] We defined 1–2 pilot use cases
[ ] We created a fitting Project
[ ] We identified 1–3 sensible Skills
[ ] Every Project and Skill has an owner
[ ] Connectors are used only on real need
[ ] Results are reviewed
[ ] Standards and versions are documented

Closing question: Which Project and which Skill should your team start with?

Schulung 3 — Claude Code & Cowork safely​

Audience: IT, Development, Security, Digital Office, power users · Duration: 25–30 min · Goal: introduce agentic functions safely.

Core message: Claude Code and Cowork are strong, but only safe with clear boundaries. Limit rights, secure environments, check results.

Learning objectives — participants:

  • understand the Claude Code security model,
  • know Cowork risks and approval limits,
  • recognize plugins, MCPs, and connectors as a governance topic.

Outline:

  1. Understanding Claude Code — agentic dev support, rights, approvals
  2. Security rules for Claude Code — review, secrets, repos, sandbox, MCPs
  3. Understanding Cowork — files, folders, deliverables, plugins
  4. Operating model & risks — pilot, permissions, approvals, reviews

The five security rules for Claude Code:

  1. No merge without human review
  2. No secrets in the prompt or context
  3. Only approved repositories
  4. Prefer dev containers / sandbox
  5. Only trusted MCPs and plugins

Participant checklist:

[ ] Claude Code is used only in approved repositories
[ ] Narrow rights are the default
[ ] No merge happens without review
[ ] Secrets are never entered
[ ] Sandbox or isolated environment is preferred
[ ] Only trusted MCPs and plugins are used
[ ] Cowork works only with approved folders and sources
[ ] Cowork results are reviewed
[ ] A pilot runs before broad rollout

Trainer note: present Cowork as a governance-relevant agentic mode, not a normal chat feature; state clearly that Cowork can be unsuitable for regulated workloads. Closing question: Which technical limit matters most for your safe start?

Schulung 4 — Configuring Claude Enterprise as admin​

Audience: IT admins, IAM, Security, platform owners · Duration: ~30 min · Goal: set up Enterprise correctly and prepare a safe go-live.

Core message: Claude Enterprise begins with identity, roles, domain, provisioning, and control — not with features. Test before you enforce. Secure before you scale.

The setup order (this order is security-critical)​

1  Check admin roles & prerequisites
2  Verify the domain
3  Set up SSO
4  Test SSO  (do not enforce yet)
5  Choose JIT or SCIM
6  Map groups and roles
7  Add audit, compliance, and network controls

If SSO or SCIM are activated too early, you get lockouts or faulty provisioning. Domain verification comes before SSO; SSO before JIT/SCIM.

Key steps and the reasoning​

  • Prerequisites: an Owner/Primary Owner, DNS access, IdP access, a check of existing Claude orgs, and a second admin as backup (avoid a single point of failure).
  • Domain verification: add the domain, set the TXT record, wait for Verified, then enable "Restrict organization creation" so employees cannot spin up shadow orgs on the company domain.
  • SSO: start "Setup SSO" in Claude, pick the IdP, complete the IdP flow, run a test login — and do not enforce immediately.
  • Provisioning: Invite only (manual) → JIT (created on first login, good for pilots) → SCIM (full provisioning/deprovisioning via IdP, the clean target state for Enterprise). Save SCIM only after all assignments are correct.
  • Groups & roles: map IdP groups to Claude roles, custom roles, and seat tiers; secure owner/admin groups; verify with a test user per group.
  • Audit, Compliance, Analytics: Audit Logs (Enterprise-only, ~180-day export, events/actor/IP/device — chat content not fully included); Compliance API for deeper governance/audit data; Analytics API for adoption and rollout steering.
  • Network controls: IP Allowlisting (approved CIDR ranges) and Tenant Restrictions (block non-approved org accounts) — test before enforcing, both can lock users out.

Verification commands (use neutral example values, never real keys or domains)​

Check the DNS TXT record:

# Linux / macOS
dig TXT _acme-challenge.example.com +short

# Windows PowerShell
nslookup -type=TXT _acme-challenge.example.com

Basic reachability check:

curl -I https://claude.ai
curl -I https://api.anthropic.com

Tenant-restriction test (the proxy injects the allowed-org header):

curl https://api.anthropic.com/v1/messages \
  -H "x-api-key: $CLAUDE_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{"model":"claude-sonnet-4-6","max_tokens":128,"messages":[{"role":"user","content":"Tenant restriction test"}]}'

Relevant proxy header: anthropic-allowed-org-ids: 550e8400-e29b-41d4-a716-446655440000

The five most common admin errors​

[ ] Enforcing SSO too early
[ ] Saving SCIM before all assignments are correct
[ ] No second admin secured
[ ] "Restrict organization creation" forgotten
[ ] Network controls enabled in production without a test

Admin go-live checklist​

[ ] Primary Owner present
[ ] Second Owner/Admin secured as backup
[ ] DNS access available
[ ] Identity Provider access available
[ ] Domain status = Verified
[ ] Restrict organization creation enabled
[ ] SSO set up, admin test login successful
[ ] Standard-user test login successful
[ ] SSO not enforced too early
[ ] Provisioning model chosen (Invite / JIT / SCIM)
[ ] SCIM saved only after all assignments correct
[ ] Group and role mapping verified
[ ] Audit Logs checked, export process documented
[ ] Compliance API enabled only on need
[ ] Analytics API enabled on reporting need
[ ] Network controls tested before enforcing
[ ] Internal operations documentation filed
[ ] Technical / IAM-Security / Data-Protection approvals obtained

Closing message: First identity, then access, then provisioning, then enforcement. Start small, test cleanly, scale afterwards.

Trainer guidance​

Before each training: check the audience, prepare 1–2 internal examples, know the contact for questions, and show no real secrets, keys, or customer data.

During: explain value and risk together, never show a feature without its governance context, involve the audience with a control question per block.

After: send the one-page handout, name the contact point, and collect pilot use cases or open risks.

For trainings 3 and 4 specifically: move slower through the risks with IT/Security groups; for the admin training, emphasize the setup order and that much of it happens in the Claude admin UI, the IdP, DNS, and proxy/network config — not via API or CLI.

Suggested rollout schedule​

Sequence the trainings with the phased rollout from the Adoption & Governance Guide:

WhenTrainingPairs with rollout phase
Before go-liveSchulung 4 (admins)Phase 1 – Governance & base
At department pilot kickoffSchulung 1 (all staff) + Schulung 2 (leads)Phase 2 – Department pilots
Before the Code/Cowork pilotsSchulung 3 (IT/power users)Phases 3–4 – Code & Cowork pilots
OngoingRefreshers + new-hire Schulung 1Phase 5 – Scale

Further reading​

All AI-generated content is a draft and requires human review before external use.