Keyboard shortcuts

Press or to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

🏠 Back to Blog

CC4E #6: The One File That Can Save Your Team Thousands of Hours

Authors: Hannah Stulberg and Joel Salinas | Published: 2026-02-21

How shared context files turn individual AI use into team intelligence.

The Problem: Re-explaining Context Is Massively Wasteful

  • Most teams use AI individually – everyone re-explains company context every session
  • The math: 5 min/session x 3 sessions/day x 100 people = 25 hours/day = 6,000+ hours/year (3 FTEs doing nothing but re-explaining)
  • Four types of context re-explained each time: company, team, role, and project

The Shift: Web-Based to Folder-Based AI

  • Web-based tools (ChatGPT, claude.ai) start every conversation blank
  • Folder-based tools (Claude Code, Cursor, Antigravity) read instruction files automatically
  • Solution: shared CLAUDE.md / AGENTS.md files

Knowledge Compounding

The core insight: when one person notices the AI making a mistake and corrects it in the shared file, everyone’s AI avoids that mistake automatically.

  • Anthropic’s own team does this – they share a single CLAUDE.md for the Claude Code repo, check it into git, contribute to it multiple times a week

Three Requirements

1. A Shared Place to Work

  • Non-technical teams: shared cloud storage (Dropbox, Google Drive, iCloud, OneDrive)
  • Technical teams: Git (GitHub, GitLab) – offers version control and change review
  • Either way, need a simple approval workflow so unapproved edits don’t slip in

2. A Folder Structure That Mirrors Your Company

Organized by departments, teams, and projects. Each level gets more specific. When a team member works on a project, all relevant context files auto-load without re-explaining anything.

3. A Habit of Contributing

The most important part is not initial setup but ongoing contributions:

  • Every time someone notices the AI making a mistake or missing context, they propose an update
  • Approval processes should match each level (team file: anyone on team; company file: leadership)
  • Over time, these become living documents capturing collective knowledge

Start Small

  • Start with one department or team
  • Create a shared folder, write a single context file together covering: what the team does, what the AI needs to know, how you want it to work
  • Use it, notice what’s missing, update. That’s the whole process
  • Once one team sees the benefit, it spreads naturally