The easiest way to make an AI follow-up system fail is to build too much of it at once. A booking platform, a CRM, an AI conversation tool, an automation layer, a WhatsApp integration, a reporting dashboard and three different tools that do slightly overlapping things, all connected by a set of workflows nobody fully understands and nobody has documented. It works in the demo. It breaks in week two. The team goes back to manual follow-up and the technology sits running in the background doing approximately nothing. The businesses that build AI follow-up systems that actually work consistently do not start with the most comprehensive possible stack. They start with the minimum viable one: the fewest tools that can reliably handle the lead, reply, qualify and book cycle, with room to add capability as the foundation proves itself. This post maps that minimum stack, explains where each component fits, names the tools that play each role most reliably and describes what the full flow looks like when it is working.