Scaling from 2 to 20

The transition from a couple of Claude enthusiasts to a team-wide deployment follows a predictable pattern, and knowing the pattern helps you navigate each stage without the common pitfalls.

Stage one: pioneer phase. One or two people on the team discover Claude and start using it for their own work. They build personal skills, refine personal prompts, and produce noticeably better or faster work. This stage is organic and requires no management. The pioneers are self-motivated.

Stage two: evangelism phase. The pioneers share their results, and 3-5 more people want to try. This is where most teams fail because the pioneers try to teach by showing their own workflows. The problem: the pioneer's workflow is optimized for the pioneer's tasks. What Bob needs is different from what the pioneer needs. Each new user needs a guided first win on their own real task, not a demo of someone else's workflow.

Stage three: infrastructure phase. With 5-10 active users, individual optimization hits a ceiling. The team needs shared infrastructure: a prompt library, shared skills, context bridge templates, consistency standards, and a governance framework. This is the phase where a team lead or operations person needs to invest time in building the shared layer. Without it, you get 10 people independently reinventing the same prompts and producing inconsistent output.

Stage four: operational phase. With shared infrastructure in place, onboarding new users becomes repeatable. The guided first win, the skill starter kit, the prompt library walkthrough, and the 30-day check-in are standardized. New hires get Claude proficiency as part of their onboarding, not as an afterthought. The team's Claude capabilities compound: every skill improvement, every prompt refinement, every new context bridge template benefits everyone.

The most common mistake at each stage is trying to skip to the next one. Teams that jump from pioneer to operational without building shared infrastructure get adoption without consistency. Teams that build infrastructure before having enough pioneers get shared resources that nobody uses because nobody contributed to them. Teams that try to enforce governance before people have experienced the value get resistance and resentment.

The sequence matters: pioneer, evangelize, build infrastructure, operate. Each stage produces the preconditions for the next stage. Pioneers produce the proof of value that motivates evangelism. Evangelism produces the user base that justifies infrastructure investment. Infrastructure produces the consistency that makes governance meaningful. Governance produces the measurement that justifies further investment. It is a flywheel, and it turns in one direction.

1. 1. Pioneer Phase (1-2 people, weeks 1-4) Let the enthusiasts explore freely. Their job is to discover what Claude is good at for your specific team. They build personal skills and produce proof of value. Management involvement: zero. Just do not block them.
2. 2. Evangelism Phase (3-5 people, weeks 4-8) Pioneers conduct guided first wins for interested teammates. Each new user starts with their own real task, not a demo. Pioneers contribute their best 3 prompts to a shared folder. This seeds the prompt library.
3. 3. Infrastructure Phase (5-10 people, weeks 8-16) Build the shared layer: prompt library, shared skills, context bridge templates, consistency standards. Assign a librarian. Create the onboarding playbook. This requires dedicated time from a team lead or ops person.
4. 4. Operational Phase (10+ people, ongoing) Onboarding is standardized. Governance is in place. Measurement is quarterly. The team's Claude capabilities compound with every skill update, prompt refinement, and new user contribution. Management reviews ROI metrics and adjusts investment.