If you’re a freelancer or small agency with a tiny (or nonexistent) sales team, you already know the pain: deals come in, the process gets messy, and you lose track of where people are at. And when you’re not on top of follow-ups, proposals, and last touchpoints, you’re basically handing deals to the next person they’re talking to. I struggle with this all the time, so in this video I walk through the agentic CRM system I built to fix it. The core idea is simple: I can ask natural language questions about my pipeline directly inside Slack—things like “What deals are stalling?”, “When did I last speak to this person?”, or “Did we send a proposal?”—and the system figures out where to look across HubSpot, Gmail, and Google Docs. Under the hood, I show the full architecture: an insights pipeline that runs daily (with a persistent database so it doesn’t repeat the same insights), and a query pipeline with a plan generator, tool registry, and a ReAct recursion framework that validates results and re-searches when needed. I also show why agent dependencies matter, how I use n8n for Slack/Gmail/Drive connectivity, and how I structure outputs with sources and a confidence score.