Anthropic has published an official "Champion Kit" teaching engineers how to get their coworkers hooked on Claude Code. It lays out three behaviors: share what you discover, be the person people ask, and grow the circle. Dev tools rarely spread through rollout announcements. They spread because someone on the team starts using them well and makes it easy for others to follow.
Fifteen to 40 minutes a week. That's all Anthropic asks for: posting screenshots, answering questions in shared channels, and running a weekly show-and-tell thread. When someone asks how you did something, hand them the actual prompt you used.
Examples from your own codebase beat external docs because colleagues can see how you debugged that flaky test or refactored that gnarly service class, problems they're wrestling with too. A concrete, runnable example closes the gap between curiosity and first successful use. That's where most adoption stalls, Anthropic says. This approach is similar to how an AI agent helped an engineer unstuck a stalled project.
It's a deliberate contrast to how GitHub pushes Copilot into enterprises. GitHub's documentation focuses on admins assigning licenses, tracking usage metrics, reminding inactive users, and establishing "AI managers." Anthropic is betting on social proof and organic habit formation instead.
Not everyone is sold. Hacker News commenters flagged the awkwardness of a corporate playbook instructing engineers to advocate authentically for a product amid rising reports of AI rollout sabotage. But the guide's genuinely well-written and practical if you think Claude Code deserves wider use on your team. Whether peers sharing screenshots beats managers mandating tooling, we'll see.