What I learn building automations for a living — Apps Script patterns, extension
engineering, and honest platform comparisons. No fluff, no affiliate links.
Where AI agents genuinely earn money in a small business — and where a one-time script beats a $50/month agent subscription. Field notes from real builds with n8n AI agents, MCP and guardrails.
The UrlFetchApp → Gemini API pattern for adding AI to Google Sheets and Gmail automations — classify, extract and draft inside your own Workspace, with copy-paste code and cost math.
Architecture for AI Chrome extensions that pass store review: side panel UI, service-worker routing, a proxy that never ships your API key, and cost control that keeps reviews short.
Custom functions, custom menus and time triggers — the three Apps Script building blocks that turn any Google Sheet into a self-running system, with copy-paste code.
Auto-labeling by content, email-to-Sheet logs, Drive attachment pipelines, follow-up reminders and data-aware auto-replies — with the Apps Script to do it.
The exact Google Apps Script pipeline I build for clients: Google Form → Sheet → Doc template → branded PDF → emailed invoice, with production-grade idempotency.
Sleeping service workers, the remote-code ban, MAIN-world injection and permission reviews — a field guide from shipping a dozen client extensions on Manifest V3.
The decision tree a working automation engineer actually uses — cost modeling, maintainability, data residency, and the failure modes comparison blogs skip.