Michael Charles had the problem every coach and agency owner wants. A video of his went viral, cleared a million views, and his Instagram and Facebook DMs flooded with people asking about their credit. He couldn’t answer everyone, so he turned on GoHighLevel Conversation AI to handle the replies.
Six weeks and roughly 800 messages later, it had booked zero calls. Within a week of switching the conversations to CloseBot, he had two booked. On the call above, he walked through what changed, and we opened his build together to make it even better. If you’re weighing the two, our HighLevel Conversation AI vs CloseBot comparison covers the feature-by-feature detail. This post is what it looks like in a real account.
Who Michael is and how leads reach him
Michael runs RSC Credit Freedom Solutions, a credit repair and funding business he’s operated for about three years. HighLevel is his CRM and his calendar. Leads come almost entirely from social… he posts, people comment a keyword or slide into his DMs, and he wants to give them a resource, learn what’s hurting their credit, and get the right people on a call.
That first message happens through HighLevel, and now the rest of the conversation and follow-up happens through CloseBot. Why use another system instead of keeping it all in-house within HighLevel? Revenue. Michael likes making money.
Michael sells three tiers
- a $77/month community for price-sensitive leads
- a $1,500 done-for-you program
- a credit complete program that includes funding
His AI setter has a job to do. It has to qualify, route, close the low-ticket offer in the DMs, and book calls for the rest. Textbook use case for conversational AI in general, and easy for CloseBot.
What 6 weeks on GoHighLevel Conversation AI looked like
The numbers were simple and brutal. About 800 outbound messages over a month and a half. Zero booked calls. Keep in mind, a lot of this depends on setup, but a lot of it doesn’t. CloseBot has been praised as being easier to set up along with having more robust features that equates to more qualified leads.
“It was sending out messages, but it had zero conversions in it.”
Michael is analytical by his own description, and he didn’t just leave it running. He spent those six weeks testing and tweaking. The problem wasn’t effort. It was that the conversations never sounded like a person, and people in the middle of a credit problem can tell when they’re talking to a script.
Week one with CloseBot: 2 booked calls
Michael found CloseBot while digging into why his GHL AI wasn’t converting. He watched three or four of our build videos, signed up, watched a few more, and started building. About three of his first seven days went to setup. Even so, the agent booked two calls that week.
“Had it for about a week… and it ended up booking two calls. So, big difference [and] follow-up is great, the message flow is great, it’s not too long. It actually feels like a human is talking to the person.”
What’s the true cost of cheaper Conversational AI? CloseBot has been an endorsed HighLevel affiliate as well as a 2x Gold Sponsor at their annual HighLevel Summit in Dallas, TX. This is not intended to be a blog that puts them down. It’s meant to show how much more money you can make in your business if you use CloseBot’s focus on Conversational AI along with your HighLevel account ❤️
60-second quiz
Which reply would you rather get?
Four real moments from a lead conversation. We’ll show you who wrote what at the end.
0/4
You picked CloseBot every time. You picked CloseBot 0 out of 4 times. Interesting. You went the other way.
Four for four. Every reply you preferred as the customer is one CloseBot sends automatically: short, natural, and written to keep the lead talking. CloseBot’s replies are designed to be natural, resulting in higher reply and conversion rates when compared to HighLevel’s Conversational AI. You picked the other reply each time. Most people go the other way, but the point holds either way: the wording of a single reply decides whether a lead answers it.
That’s the core of the gap between a general-purpose GHL AI chatbot and a dedicated AI appointment setter. CloseBot splits replies into short, natural texts, times them like a person would, and is built around one outcome: getting the conversation to a booking.
Inside the build: one HighLevel AI agent, routed by intent
When we opened Michael’s account, there wasn’t much to fix. He had built it with Claude, talking through his idea and having it set things up through the API, and he’d done his homework on our videos.
His first agent node does the qualifying: what’s hurting your credit, how long has it been going on, where do you want your score to be. It saves those answers straight into HighLevel custom fields (main credit issue, credit score, credit goal). Then it routes the lead to an Academy path or a booking path, each with a lighter prompt, which keeps the cost per message down after the handoff.
Two details stood out:
- Social proof on demand. If a lead mentions a collection or a late payment, the agent sends an image of that exact item being removed from a past client’s report.
- Smart FAQ plus booking tools on every node: check availability, book, cancel, and reschedule, all writing to his HighLevel calendars.
“I didn’t have the right tool, and this definitely is the right tool, for sure.”
The tweaks we made on the call
Most of the session was polish, and all of it is useful for anyone running a GoHighLevel AI booking bot over DMs:
- Follow-up states that actually switch. Michael had a warm follow-up (30 minutes, 4 hours, 20 hours) plus Academy and Booking cadences. The agent was never told when to switch, so everyone stayed in warm. We added the instruction to move to the Academy cadence after a clear yes.
- Knowing when to stop. If a lead gets upset or asks to stop, the agent switches follow-ups off. Michael had already tested this by messaging aggressively. The agent recognized it, did a human handoff, and flagged him to jump in, which he likes because a quick voice note softens things.
- Not replying when a reply would be weird. A liked message or a goodbye doesn’t need a “glad you liked that!” The agent can choose silence, which reads as more human.
- GIFs in follow-ups. We turned on the GIF tool so follow-ups can send the playful “you there?” GIF Michael wanted.
- SMS for the 24-hour window. Instagram and Facebook only allow replies for 24 hours after the lead’s last message. Collecting a phone number lets the conversation move to text and keep following up long term. Michael plans to add SMS for referral leads from his mortgage lender partners too.
And if Michael jumps into a thread himself, CloseBot adds an AI off tag. Removing the tag hands the conversation back, and the agent picks up where it left off.
What it costs and what he can see
CloseBot charges a flat 1.2 cent per message segment unless you unlock unlimited prompt size on an agent node. Michael’s first week came to about 200 segments, roughly $2. Compare that to how GoHighLevel Conversation AI pricing works in our pricing breakdown.
We also built him a daily pulse dashboard in CloseBot Analytics: his funnel (contacts, responded, booked, in follow-up), bookings per week by channel, wallet spend over time, topics with sentiment, and a heat map of when leads message. His first week showed Saturday evenings and weekday middays as hot spots, which tells him when to be around for a voice note.
Is GoHighLevel Conversation AI enough?
For a lot of HighLevel users, native AI is fine for answering basic questions. But if you’re counting on it to convert traffic into calls and into revenue, CloseBot’s impact is worth it.
Michael’s experience shows where it falls short… when the job is to qualify, route, follow up, and book, natural conversation is the whole game. Six weeks of testing didn’t move his GHL AI off zero. One week on a purpose-built setter did.
“If I got all those areas, man, I’ll be a dangerous man.”
See the full HighLevel vs CloseBot comparison, or start CloseBot free and connect it to your HighLevel account in minutes.
Frequently asked questions
Chatbot analytics is the measurement of what an AI chatbot or AI sales agent does and what it produces. That covers conversation volume, engagement per contact, sentiment, bookings, lead scores, and cost. For sales use cases, the most useful chatbot analytics connect AI activity to outcomes, such as cost per booking.
Yes. CloseBot connects to your HighLevel account, books directly onto your HighLevel calendars, and writes what it learns into your custom fields and tags. HighLevel stays your CRM; CloseBot runs the conversation.
Yes. CloseBot replies through the channels your HighLevel account already has, including Instagram and Facebook Messenger. Because Meta only allows replies within 24 hours of a lead’s last message, the agent can collect a phone number and continue the conversation over SMS.
HighLevel’s native AI is one feature inside an all-in-one platform. CloseBot is built only for sales conversations: multi-bubble human-style replies, follow-up states the AI switches between, a testing portal, Smart FAQ, and conversation-level analytics. See the full HighLevel vs CloseBot comparison.
No. The AI Employee is HighLevel’s bundle of native AI features, including Conversation AI. CloseBot is a separate platform that replaces the conversation layer while working inside your HighLevel account.
It depends on your lead volume and offer. Michael spent about three days building his agent and booked two calls within his first week.
CloseBot charges a flat 1 cent per message segment unless you unlock unlimited prompt size on an agent node, which switches that node to token-based cost. Michael’s first week was about 200 segments, roughly $2. Our pricing comparison breaks down HighLevel’s usage-based and unlimited options.
Yes. When you send a manual message, CloseBot can add an AI off tag to the contact. Remove the tag and the agent picks the conversation back up where it left off.
No, if you tell it not to. You can instruct the agent to turn follow-ups off when a lead is upset or asks to stop, and to hand the conversation to a human. It can also choose not to reply to things like a liked message or a goodbye.
Yes. Michael built his agent by talking his plan through with Claude and having it set things up through the CloseBot API. A CloseBot MCP server is launching to make that even easier.