AI sales agent in GoHighLevel cover: Seam 13, The Loop, from The Thirteen Seams by Jon Weberg

Calls Booked While You Sleep: An AI Sales Agent in GoHighLevel

October 26, 2026•15 min read

Open your GoHighLevel sub-account, click AI Agents, and read the word beside your bot's name, because whether it says Off, Suggestive or Auto-Pilot, the worker you pay for every month is still waiting for a job. Somewhere in last week's Conversations tab is a lead who texted at 9:40 on a Tuesday night asking whether Thursday had an opening, and you know how many hours she waited, don't you? The gap between that bot and one that books her Thursday is the gap between a hire handed a login and one handed a written role, a boundary, a handoff and a number for Friday.

The short answer

An AI sales agent in GoHighLevel is a Conversation AI bot or Voice AI agent, sold in HighLevel's AI Employee plan, that you hire for one written job: a single handoff such as first reply, booking or show confirmation. You give it a knowledge base, a Human Handover rule for the moment it is unsure, workflow handoff rules and one number, then test it for 7 days before full Auto-Pilot.

What you will walk away with

  • A six line job description (job, seam, inputs, boundary, handoff, number) written before you open the bot builder.

  • The exact click path through AI Agents, Conversation AI and Bot Goals, with the worker running as a Non-Primary bot started by a workflow.

  • All three Human Handover triggers switched on and routed to a named person with a task due inside 24 hours.

  • A 7 day test, two days in Suggestive mode and four in Auto-Pilot on one channel, judged on one number against your own previous 7 days.

  • The pause rule your setters need, because in Auto-Pilot a human reply alone does not stop the bot.

What is an AI sales agent in GoHighLevel, and what does the AI Employee plan give you?

An AI sales agent in GoHighLevel is any of the HighLevel AI products that talk to your leads, and seeing them as separate workers lets you hire each for the job it does best. Conversation AI answers in writing across SMS, Email, Facebook, Instagram, WhatsApp, the Chat Widget and Live Chat, per HighLevel's guide on how to set up a Conversation AI bot. Voice AI answers and places calls, books into your calendars and hands the caller to a live person, running only on LC Phone or Twilio numbers, per the AI Voice Agents overview. Workflow AI adds actions such as AI Decision Maker and AI Intent Detection to your workflows, and the Knowledge Base under AI Agents holds what your agents are allowed to know.

With AI Employee Unlimited listed at $97 a month per location in HighLevel's AI Employee pricing article, your least expensive hire is already in your account, and it deserves the written role you would hand a new setter, doesn't it?

Why does a bot with a channel but no job never book the calls you bought it for?

A bot misses bookings because it was hired into a channel instead of a handoff, which was never your prompt or your model. Marketing owns the form, setters the dials, closers the calls, and the moment between each pair belongs to nobody, so pointing Conversation AI at SMS hands a worker every conversation on a channel while the handoff it was meant to close still has no owner, no rate and nobody waiting on the other side.

"That's kind of everyone's job." "We looked at that last year and it stalled." "I end up doing it myself."

If you said the third, you know who has been holding the joints together at 10pm. When I built the custom CRM that AI Acquisition, a $70M a year company, runs on, the lead routing, what reps tell leads and what they say on calls each had a written owner and a rule for where the work went next, because a system built on a broken handoff simply runs the leak faster, and an AI worker is a system that types quickly.

This is Seam 13, THE LOOP, in The Thirteen Seams: an AI worker given one handoff to own, with its first jobs on Seam 03, CAPTURE to CONTACT, Seam 04, CONTACT to BOOKING and Seam 05, BOOKING to SHOW.

Which handoff should your first AI sales agent own?

Your first AI sales agent should own whichever of Seams 03, 04 and 05 costs you the most, and three questions answered from your own CRM tell you which. How many of last month's form fills got a first reply inside five minutes that mentioned something they typed? The InsideSales.com and MIT Lead Response Management Study (2007, over 15,000 web leads) found the odds of contacting a lead drop 100 times between a call at 5 minutes and one at 30, and the odds of qualifying drop 21 times, which is the case for an AI worker on speed to lead at Seam 03.

How many leads who said "not yet" ninety days ago have heard from you since? That is Seam 04, where an agent runs your lead follow up sequence or a lead reactivation campaign on pools your team wrote off. And who speaks to a prospect between booking and the call, and what share of last week's bookings appeared? That is Seam 05, where an agent that confirms and asks for a reply can increase show rate before your closer dials.

A worker only carries a handoff a person has already closed by hand, because an agent cannot invent a routing rule your team never agreed on, which is the order the Thirteen Seams framework follows. On John Crestani's list, each group got its own sequence plus a text channel, and the list paid $10,000 to $30,000 a week for 15 straight weeks, because every group's next step was written first. To see your three costliest handoffs priced on your own numbers, book a free Seam Map Review.

How do you write the job description for an AI worker?

You write an AI worker's job description in six lines, the same six you would want on a setter's first day: the job, the seam it owns, the inputs it reads, the boundary where it stops, the handoff that names who takes over, and the one number it answers for on Friday.

AI worker

Job

Owns which seam

Inputs

Boundary

Handoff

Number

First reply agent

Text every new lead back, opening with what they told you

Seam 03, CAPTURE to CONTACT

Form answers, offer pages, objection doc

Any price or eligibility question not in the knowledge base

Human Handover to the assigned setter, task due in 24 hours

Share of new leads answered inside 5 minutes with a reference to their form

Booking agent

Turn a reply into a booked call

Seam 04, CONTACT to BOOKING

Calendar, qualification questions, past messages

Lead asks for a person or fails qualification

Booked: workflow to the closer. Not a fit: tag to the smaller offer

Share of replying leads who book

Show agent

Confirm every booking and get a reply

Seam 05, BOOKING to SHOW

Appointment time, closer name, what the call delivers

Reschedule request outside open slots

No reply 2 hours before the call: task for the closer to dial

Share of booked calls that show

Reactivation agent

Reopen dormant leads with one dated reason to talk

Seam 04, CONTACT to BOOKING

Smart list, last note, original offer

Any question, complaint or "call me"

Reply pauses the bot and creates a call task for the setter

Booked calls per 100 dormant leads contacted

Every boundary is a moment, never a mood, so the worker can be judged in a week instead of argued about for a quarter, can't it?

How do you set up Conversation AI in GoHighLevel so it answers like your best setter?

You set up Conversation AI by building the bot around the job description, in this order, using HighLevel's current names.

  1. Go to AI Agents, Conversation AI, + Create Bot, and pick Prompt Based Bot with the Appointment Booking type, or the Flow Based Builder when you want the Stop bot, Human handover and Auto follow-up actions drawn as a visual flow, per the bot setup guide.

  2. In Bot Settings, choose Non-Primary for a worker that owns one seam: a Primary bot handles general inbound on its channels, while a Non-Primary bot responds only when a workflow invokes it, per the Conversation AI bot explainer, so the workflow that marks the start of your seam hires the worker for each lead.

  3. In Bot Training, feed it your words: the Web Crawler (Exact URLs, All URLs with Path or All URLs in Domain, then Get Data), a Google Doc of objections shared as Anyone with the Link, and your real answers through + Add Q&A as Custom Bot Responses, per Training Your Conversation AI Bot. A Knowledge Base trains up to a fixed number of web URLs, such as 4,000, per HighLevel's Knowledge Base web crawler article.

  4. In Bot Goals, set the questions to ask, booking behavior and workflow triggers, paste your job, boundary and handoff lines into Additional Instructions, and pick a Brand Voice and a Concise, Balanced or Detailed response style.

  5. Set "Maximum messages a Bot can send in a Conversation", which runs from 1 to 100 and defaults to 100, per HighLevel's Auto-Pilot mode article, to the number of messages your best setter would need for this one job.

  6. Test from the setup panel: FAQs, booking, a question it should not know, and every workflow it fires.

How does your AI agent know to call a human the moment it is unsure?

Your AI agent calls a human through the Human Handover action in Bot Goals, under Setup Your Actions, and HighLevel's Human Handover article lists three triggers: Human Requested, when the lead asks for a person; Lack of Information, when the AI cannot find an answer; and Failed to Resolve Issue, after repeated failed attempts. On handover you set Assign Conversation to a User, a Final Message, a Bot Pause Behaviour timer, Create a Task due 24 hours later and Create Tags (default human_handover), while each setter turns on "When a conversation gets assigned to me" and "When a task gets assigned to me" in Notification Settings so the handoff rings a phone.

In Auto-Pilot a human reply does not stop the bot, which keeps answering until its message cap or a pause, and marking the conversation as Read wakes it and resets its counter, per HighLevel's Auto-Pilot article. So the setter rule is written before day one: when you take a conversation, a workflow running Update Conversation AI Bot and Status sets that contact's bot to Inactive, and because the bot sends no goodbye, your Final Message tells the lead a person is on it.

On a call the same boundary is Call Transfer to a live person's number, per the Voice AI Agent Transfer article.

How do Workflow AI actions route every lead back to the right person?

Workflow AI actions route each lead through branches you write in plain English, so every lead ends at a named person or step rather than in a thread nobody opens. AI Decision Maker reads your instructions plus the contact's fields and tags and picks a branch you named, with a locked Default Branch catching the rest, per the AI Decision Maker article, while AI Intent Detection sorts a reply into POSITIVE, NEGATIVE or NONE at one cent per execution, per the AI Intent Detection article.

From a real install

When a tax strategy firm wanted the dormant leads in one closer's pipeline worked again, I put 1,935 of them on a single smart list that excluded anyone already booked, bought, opted out or spoken to in the last 14 days, with every handoff rule written before the first message. A setter texts a heads up 2 to 3 minutes before dialing from the closer's own number; on no answer an automated cadence sends a Day 3 text ("Still worth a look ... before the deadline, or should I close your file out?") and a Day 7 final text with the closer's calendar link, then hands the lead to a 21 day nurture on Day 8. Any reply instantly stops the cadence and creates a CALL NOW task for the setter, a booking stops it, "stop" sets do not disturb on every channel, and leads who are not a fit for the main offer go to a $997 smaller offer with its own two email sequence, Email 2 sending only if Email 1 got no click, with texts going out Monday to Saturday, 10am to 6pm in the lead's time zone. That is the shape of every AI worker you will hire: one list, one job, a boundary that stops the machine when a human is needed, and a named person waiting.

How do you run a 7 day test, and which one number decides if the AI worker stays?

You run the 7 day test like a new setter's first week, with your previous 7 days as the baseline and the job's one number as the only score.

Day

Bot status

What you do

What you read

Day 1

Off

Write the six line job, load training, test every path

Your seam's rate for the previous 7 days

Days 2 and 3

Suggestive

The bot drafts, your setter edits and sends each reply

Drafts sent unedited versus rewritten, and why

Day 4

Auto-Pilot, SMS only

Low message cap, all three handover triggers on

Every conversation tagged human_handover

Days 5 and 6

Auto-Pilot

Fix Additional Instructions and Custom Bot Responses from those tags

Dashboard: unique contacts, actions triggered, appointments booked

Day 7

Auto-Pilot

Compare the one number with your baseline week

Keep, rewrite the job, or move the worker to another seam

Suggestive mode drafts replies for review while Auto-Pilot sends them alone, per the Conversation AI explainer, so the first live days show you where your knowledge base is thin. By Day 7 you are reading one rate beside last week's with a name beside both, which is the conversation you have wanted about every handoff, isn't it?

Questions owners ask about AI sales agents in GoHighLevel

Is Conversation AI included in the GoHighLevel AI Employee plan?

Yes. HighLevel's pricing article lists AI Employee Growth at $50 a month per location with 1,000 Conversation AI agent responses and 100 Voice AI minutes, and AI Employee Unlimited at $97 a month per location with Conversation AI and Voice AI unlimited, subject to fair use. Agent Studio stays pay per use, and phone system charges apply on every plan.

What is the difference between Suggestive mode and Auto-Pilot in GoHighLevel?

Suggestive mode has the bot draft a reply that a person on your team edits and sends by hand, while Auto-Pilot sends replies on its own based on its training, goals and instructions. Off means the bot does not respond. Run Suggestive first, so every draft your setter rewrites becomes a training fix before the worker sends anything alone.

Can GoHighLevel Voice AI call my leads outbound?

Yes, through the Voice AI Outbound Call workflow action, which needs an AI Agent and a From Phone Number. HighLevel caps it at 10 calls a minute and 5,000 a day per location, once a day and up to 14 times in 14 days per number, between 8:00 AM and 8:00 PM in the contact's time zone. Voice AI runs on LC Phone or Twilio numbers.

How do I stop the GoHighLevel bot when my setter takes over a conversation?

Use a Human Handover with a Bot Pause Behaviour timer, or a workflow with Update Conversation AI Bot and Status set to Inactive for that contact. A human reply alone does not stop a bot in Auto-Pilot, and marking a conversation as Read wakes the bot and resets its counter, so write the pause rule for your setters before launch.

Before this, your AI Employee was a line on the invoice with a channel and no job, and right now you are one written page away from a worker whose handoff, boundary and number are on paper, so a week from Monday you are reading one rate beside last week's and deciding calmly. Bring the numbers you already have to your free Seam Map Review, where I walk all thirteen seams with you live and put a monthly dollar figure beside your three largest, and if you want the full hiring method tonight, start with INSTALL, the $7 book on hiring AI workers.

blog author avatar

Jon Weberg

Jon Weberg finds and closes the seams between your teams through Profitalize, by installing an AI Seam Architect, by consulting beside your team every month, or by running your AI OS, and every part of it is built inside your team, your offer and your accounts.

Back to Blog
Jon Weberg, founder of Profitalize
Written by

Jon Weberg

Founder, Profitalize · AI growth seam architect

Jon finds the handoffs where revenue leaks between your teams and installs the AI workers that own them. As Fractional President of AI Acquisition he directed its Sales, Marketing and Media Buying Directors and built the CRM a $70M a year company runs on. He is a two time published author, an international keynote speaker at Affiliate World, Traffic and Conversion Summit and Affiliate Summit, and has spent 13 years in direct response.

  • $266,500 in 7 days · Tax Strategists of America
  • $41,253 in 3 days · Frank Kern
  • $270M referral revenue · NowLifestyle

About JonCase studiesLinkedInYouTube

Your next step

Find your first seam this week.

Forty five minutes on your own numbers, or the $7 book that has your first AI worker running by Friday.

Free. Forty five minutes. Your seams, priced, before we hang up.