Systems & InfrastructureAutomationsCRM Architecture

GHL System: twenty two workflows, a flow AI agent, a setter briefing AI agent, and an automated setter queue

Twenty two workflows inside GoHighLevel. A flow AI agent that holds a real conversation, knows the treatments and the prices, offers real times out of the live diary and books one. A setter system where the queue builds itself: the setter presses start and works it, and every call arrives with the notification that raised it and an AI summary of the conversation already on the record. Nine pipeline stages behind all of it. Every workflow is on this page and you can open any of them and look inside.

Two systems running at once. On the left, what happens on its own: down the middle when someone books, off to the sides when they do not. On the right, the setter system, where the queue builds itself and the call ends in one tap.Open any of these workflows

The problem

Most clinics do not have a lead problem. They have a conversion problem. The leads arrive and then they sit: nobody answers at nine at night, nobody chases the ones who went quiet, nobody rings the no shows, and by Friday nobody can say who actually turned up. I was asked to close that gap end to end, and to build the setter team into the system rather than beside it. What came out is twenty two workflows inside GoHighLevel, an AI that talks to people and books them, a queue that dials for the setters, and an agent that briefs the setter on who they are about to call.

What shipped

Twenty two workflows that install as one package, rather than being rebuilt for each clinic
A new lead gets an answer in seconds, at any hour of the day
The clinic records who turned up with one tap in an email, no login

What happens

Someone sees an ad and fills in a form. Within seconds they get a message from the clinic asking what made them look into it, and offering to book them in. If they reply, a bot answers them properly: it knows the treatments, it knows the prices, and it can offer real times from the real diary and book one.

If they go quiet, they get followed up for a week and then left alone. If they book, they get a confirmation, a reminder the day before, and a reminder two hours before. After the visit the clinic gets an email with two buttons, showed or no show, and one tap records it. A no show gets chased for a new time. Someone who turned up gets asked for a review, once. If they are partway through a course of treatment, the system keeps track until it is finished.

Alongside all of that, the phone. Anyone the messages have not booked goes onto a call list on their own: the new lead who has not replied, the person who saw times and stalled, the cancellation, the no show. The setter does not build the list, pick who to ring, or type a number. They press start and work the queue, and every task arrives with the notification that raised it and an AI summary of the conversation already written onto the record, so nobody reads back through a thread before dialling. When the call ends they tap one outcome and the system does the rest: moves the record, stops the right chases, books the callback on the day it was promised.

None of that needs anyone at the clinic to remember to do anything.

The pipeline

The workflows

Custom fields, custom tags, custom values

Inside this system