Case Study: Reducing Firestore reads from 151k/week to 8 in an AI Executive Assistant

I recently built a solo-business AI Executive Assistant using a “Board of Directors” stack (CFO, COO, CMO, etc.).

The biggest win: I was hitting 151,000 Firestore reads per week due to polling. By refactoring the local worker to use onSnapshot, I brought that down to 8 reads per week with no loss in performance.

Full architecture and sanitized code available here: AI-Executive-Assistant Repo