Onboard Your Whole Team To Claude In 30 Days

You spent months building a Claude setup that runs your workflow, but your team still starts from zero. Anthropic's new /team-onboarding command generates your team SOP in 2 minutes. The harder problem is getting your team to actually use it, and this playbook covers both layers.

What You'll Learn

• How /team-onboarding works: the 2 data layers it reads and the 6-section SOP it produces
• The CLAUDE.md architecture that turns thin output into a production-grade team SOP
• The harder problem: how to actually get your team to ADOPT the SOP (the part nobody talks about)
• The 30-day team enablement playbook: week-by-week how to take your team from 0 to 10x with Claude
• Failure modes, technical (Claude Code) and adoption (team behavior), and the fix for each

The Real Problem You're Trying To Solve

Every business owner has the same arc. They personally spend 3 months building a Claude setup that runs their entire workflow. Custom skills. Slash commands. CLAUDE.md. MCP servers. The system runs. They feel like they have superpowers.

Then they try to get their team to use AI. The team starts from 0. They don't know the prompts. They don't know the workflows. They don't even know what's possible. The owner has to drop their work to walk through every task. After 2 or 3 attempts, the team quietly stops asking and goes back to doing things the old way. The AI "transformation" becomes a one-person show.

The root cause: solo AI workflows are tacit knowledge. Every decision lives in the operator's head: the prompt that worked, the workflow they refined over 50 iterations, the gotchas they learned the hard way. None of it is documented. When the team is handed the tool without the tacit knowledge, they get the worst version of AI (slow, generic, frustrating) and conclude AI doesn't work for their job.

Every team that fails at AI fails for one of two reasons. Either the technical setup is wrong (no shared CLAUDE.md, no documented skills, no SOP) OR the adoption is wrong (the SOP exists but nobody uses it). /team-onboarding solves the first problem in 2 minutes. The second problem is on you. This lesson tackles both.

/team-onboarding: The Technical Unlock

Anthropic shipped /team-onboarding in Claude Code v2.1.101. It's the first command that solves the tacit-knowledge problem in 2 minutes. Run it once at the root of your Claude Code project. It reads two data sources, cross-references them, and writes a complete team SOP.

Layer 1: structural data (what exists). Your .claude/commands/ folder. Your .claude/skills/ folder. Your .claude/agents/ folder. Your CLAUDE.md file. Your MCP server registrations. Any sibling repos the command auto-detects. This is the static snapshot of what's available.

Layer 2: behavioral data (what you actually use). Your last 30 days of Claude Code session history. Which slash commands fired, how often. Which work types those commands map to. Which MCP servers you actually touched vs the ones registered and ignored. This is the dynamic layer: what you actually reach for.

The cross-reference is the unlock. Any tool can list files. Very few can tell a new teammate that the senior operator runs /claude-scout 7 times a week and has never touched /migrate-schema. That second layer turns a file listing into a real SOP. It compresses "here's everything we have" down to "here's what actually matters in your first week."

The 6-Section SOP Output

The generated markdown follows a consistent 6-section structure. It functions as a ramp-up guide on day one AND a living SOP your team references long-term.

1. Work type breakdown. Percentage breakdown of your time across categories: writing docs, building features, planning, debugging, reviewing, refactoring. Computed from your last 30 days of slash command usage.
2. Top commands ranked by frequency. The 5 to 10 slash commands the senior operator runs constantly. New teammates learn these first instead of trying to learn the entire library. Highest-leverage section in the output.
3. Custom skills plus MCP servers documented. Every tool the project depends on, listed with its purpose.
4. Architecture notes pulled from CLAUDE.md. The WHY behind your design choices. Constraints. Decisions made and rejected. This section is ONLY good if your CLAUDE.md is good.
5. First-day workflow for new teammates. A specific sequence: read this, run that command, look at this output, expect this behavior. Concrete onboarding, not abstract documentation.
6. Sibling repos and project paths. Related codebases the command auto-detected from your usage patterns.

The CLAUDE.md Lever (Why The First Run Is Usually Mediocre)

The first time most people run /team-onboarding the output is thin, usually a 2 to 3 page file that reads like a feature list. This is not a bug in the command. It's a signal that your CLAUDE.md is shallow. The output is downstream of the input. If your CLAUDE.md is 8 lines, the generated Architecture Notes section will be 8 lines.

The fix is upstream: before you run /team-onboarding for real, spend 2 to 4 hours rewriting your CLAUDE.md into a real architecture document. Document the WHY. Document the constraints. Document the decisions you made and the ones you rejected. Then re-run /team-onboarding. The output jumps from 2 to 3 pages of mediocre to 10 to 15 pages of production-grade. Same command, way better input.

The CLAUDE.md skeleton that 10x's your output:

1. Team / Project Identity
   - What this project is, who owns it, who maintains it
2. Architecture Principles
   - The 3-5 non-negotiable rules you've adopted
   - The patterns you use and the ones you reject (with WHY)
3. Decision History
   - The 5-10 biggest decisions made on this project
   - Each one: what we chose, what we rejected, why
4. Skill / Command Conventions
   - Naming rules for custom skills + slash commands
   - When to write a new skill vs use an existing one
5. Failure Modes + Fixes
   - The 5-10 things that have broken on this project
   - How each was fixed
6. Integration Notes
   - Which MCP servers, why, and which ones are flaky
   - Auth + secret-management conventions
7. Team Workflows
   - How AI fits into your PR/review/ship cadence
8. Things I Wish I Knew On Day 1
   - Pure tribal knowledge, written down for once

The Harder Problem: Getting Your Team To Actually Use The SOP

/team-onboarding generates a beautiful SOP in 2 minutes. Whether your team USES it is a completely different problem. Most teams don't, and the SOP becomes another doc in another folder nobody opens.

Why teams reject the SOP: it is asking them to change behavior. Not "read this doc." Actually use Claude where they currently use Google plus Slack plus their own head. Behavior change is the hardest thing in any team. Without a deliberate adoption plan, the SOP gets shared once, scanned by 2 people, and forgotten by Friday.

The 5 adoption blockers and how to beat them:

1. "I don't know what to use Claude for." Fix: pair each teammate with 1 of the top commands from Section 2 of the SOP. One command, one week. They learn by doing, not reading.
2. "Claude gave me a bad answer once, it's not for me." Fix: weekly 30-min team session where you debug bad outputs together. Bad outputs are CLAUDE.md gaps, not Claude failures.
3. "I don't have time to learn this." Fix: replace one existing task they hate with a Claude workflow that does it in 1/10 the time. The time saving is the proof.
4. "My manager doesn't expect me to use this." Fix: tie SOP usage to a visible metric (weekly check-in: "what did Claude do for you this week").
5. "I forgot the prompts." Fix: the SOP itself is the answer. Pin it. Print it. Reference it in every team meeting for the first month.

The 30-Day Team Enablement Playbook

Run /team-onboarding to generate the SOP. Then follow this 4-week sequence to actually get adoption.

Week 1: set the table. Generate the SOP. Improve CLAUDE.md. Re-generate. Share the SOP with the team in a single 30-min meeting (not async). Frame it as "here is how we use Claude, ask anything." Pair each teammate with 1 specific command from the Top Commands section.

Week 2: forced practice. Each teammate uses their assigned command at least 3 times. End-of-week 15-min standup: each person shares one good output, one bad output. Bad outputs become CLAUDE.md gaps to fix.

Week 3: real work. Move from forced practice to real-work integration. Each teammate identifies one current task they're going to replace with a Claude workflow this week. You shadow and help the first time. After that they're on their own.

Week 4: make it permanent. Re-run /team-onboarding to capture the team's actual usage over the last 30 days. Compare to week 0. The delta is your enablement ROI. Share it back to the team. Set the monthly cadence going forward.

The metric that matters: track weekly Claude usage per teammate. Not perfect attendance. Not number of prompts. Just: did each person use Claude for real work this week. After week 4, you should see 70%+ of the team using Claude on real tasks weekly. If you're below 50%, the SOP is fine and the adoption plan is failing. Iterate the adoption plan, not the SOP.

When To Re-Run /team-onboarding

The 4 trigger moments:

1. The day before a new teammate starts. The output is freshest then.
2. Monthly on a fixed cadence (last Friday of the month). The versioned snapshots become your team's SOP version history.
3. After you add 3+ new skills or slash commands. The SOP needs to reflect the new shape.
4. Before any retro, planning session, or quarterly review. A current SOP changes the conversation.

Failure Modes (Technical Plus Adoption)

• Technical: output is thin. Your CLAUDE.md is shallow. Spend 4 hours rewriting CLAUDE.md using the skeleton above, then re-run. The single highest-leverage fix.
• Technical: output is too long / overwhelming. You have too many skills and commands and the SOP reads like a catalog. Add a "Start Here" section to your CLAUDE.md pointing at the 3 commands that matter most. /team-onboarding will surface them at the top of the output.
• Adoption: team read it once and never came back. You skipped the forced-practice week. Reset. Pick one command per teammate. Make them use it 3x this week. Real adoption requires reps, not reading.
• Adoption: only 1 or 2 people use it, the rest opted out. The early adopters are doing the proof-of-concept. Get them to share specific outputs in the next team meeting. Other teammates need to SEE the value, not hear about it. Visible proof is the unlock.

What To Do This Week

1. Open your most-used Claude Code project
2. Make sure you're on Claude Code 2.1.101+
3. Run /team-onboarding
4. Read the output and notice what's thin
5. Rewrite CLAUDE.md using the skeleton above (2 to 4 hours)
6. Re-run /team-onboarding and save the production-grade output
7. Schedule the Week 1 team meeting from the playbook
8. Run the 30-day adoption sequence
9. Re-run /team-onboarding at the end of week 4 to measure

Go Deeper

• Printable PDF version of this guide: https://drive.google.com/file/d/1qoAAfvSzDtj1adUVCF6OBfRjwNr2BTG6/view?usp=drivesdk

Want the whole system installed live, with Cooper on the call? Build Your Own AI, the two week sprint runs the first two weeks of August.

Or get a new working tool every Monday inside The AI Operator community on Whop.