Building a Private AI for Your Company: Phase 0 Testing

Follow along as I start building a private AI system for a company. This is phase zero: before building anything for real, we test that masking private data, routing, and LibreChat will actually work. I already did a lot of planning, so I clear the session and write one prompt for Claude Code that points to all of it: the plan, the progress page, and the style guide. Claude reads it, makes its own plan for phase zero, and asks me to okay a few things like downloads, a clean virtual environment, LibreChat traffic capture, and an Azure account for Microsoft 365 sign-in. I also show the "by the way" feature, which lets you ask a side question while Claude keeps working. The build ran for over an hour with background agents, and Docker needs a reboot to finish, so that will pick up in the next video. This is for business owners and managers who want to see what it actually takes to set up AI that keeps company data private. 0:00 Where the project stands 0:42 Writing the phase 0 prompt 2:29 Claude makes its own plan 3:20 What phase 0 tests and why 3:40 Approving downloads and Azure 5:04 Progress page updates 5:16 Using a by the way question 6:02 Long running build and next steps I help small and mid size companies integrate AI into their workforce safely and privately. Want help with your team? https://dancdyer.com/ai-consulting Email: dan@dancdyer.com