What we learned after giving every employee a Claude Code account and forcing them to use it for 6 months

Our in-house CC Analytics — how we track each teammate's Claude Code usage.

- The position most enthusiastic about this tool was PM

1) Seems to stem from a kind of liberation — no longer having to depend on developers (ranging freely across frontend, backend, design, devops — like Jayden Jeong)

2) Conversely, response from developers was relatively low — reminiscent of the Luddite movement (though when a team member uses it well, their impact is the most explosive)

- As AI natives, the KPI most closely tied to growth was the number of prompts entered per day.

1) Similar to how, in weight training, the "go heavy" camp beats the "lift pink dumbbells and focus on the mind-muscle stimulus" camp. In the early stage, don't worry about how well you use it — volume is what matters.

2) Other KPIs with big impact: the proportion of prompts that were questions; frequency of using words/phrases like "root cause," "explain me easy," "why"; and session ratio (prompts per session — how deep you dig in).

- The gap between the top 5% and bottom 95% is really large, and this gap tends to grow like compound interest over time.

- Traits once undervalued — knowing things broadly but shallowly, multitasking, ADHD, being a mania for new tools — are now standing out in the AI era.