How to Set Up AI in GoHighLevel (Conversation AI, Its Limits and Alternatives)

There are two ways to add AI to GoHighLevel: HighLevel's built-in Conversation AI (part of its "AI Employee" suite), or an external AI setter that connects to your sub-accounts and replies through GHL's conversations. The built-in option is quick to switch on and fine for FAQs and simple booking. External setters usually add a structured sales process, deeper CRM actions, testing tools and more control over the AI model.

I built ReplySetter's GoHighLevel integration, so I know GHL's conversation APIs well, and I've set up Conversation AI to compare. I'll cover both fairly. HighLevel updates its AI features often, so check their current docs for exact menus and pricing.

Option 1: GoHighLevel's built-in Conversation AI

What it is

Conversation AI is HighLevel's native AI bot. It works inside a sub-account and can reply on the channels connected there: SMS, Facebook, Instagram, WhatsApp, live chat and others. It can answer questions from a knowledge base you give it (FAQs, website content) and can book appointments on GHL calendars.

How to set it up (general steps)

1. In your sub-account, open the AI Agents / Conversation AI settings.
2. Create a bot and give it a name, a goal (e.g. book appointments) and a personality.
3. Add knowledge: your website, FAQs and documents.
4. Choose channels it should respond on.
5. Pick a mode: suggestive (it suggests replies for staff) or auto-pilot (it replies on its own).
6. Connect a calendar if you want it to book.
7. Test in the built-in tester, then turn it on.
8. Use workflows to trigger actions based on what happens in conversations.

Where it works well

• FAQ answering on a single sub-account.
• Simple booking flows.
• Teams that want everything in one HighLevel bill.
• Quick setup with no extra tools.

Where I've seen it fall short for appointment setting

From my testing and from agencies who've moved to external setters, the common complaints are:

• Limited sales structure. It's harder to build a multi-step qualification process where each step has its own goals and exits.
• Visibility. When it says something odd, it can be hard to see exactly why.
• Model choice. You use the models HighLevel provides.
• Testing and A/B tests. Limited ways to compare two versions of a script.
• Multi-account agencies. Setting up and maintaining a bot per sub-account takes time.

None of these matter if your needs are simple. They matter a lot if setting is your business.

Option 2: An external AI setter connected to GoHighLevel

How the connection works

External setters connect to GHL in one of two ways:

• A Private Integration token for a single sub-account.
• Marketplace OAuth to connect many sub-accounts at once (common for agencies).

Once connected, inbound messages arrive by webhook. The setter reads the conversation and contact, writes a reply, and sends it back through GHL on the same channel (SMS, WhatsApp, Instagram, Facebook, email or live chat). It can also update contact fields, tags, opportunities, notes and tasks, add contacts to workflows, and book on GHL calendars.

What you typically gain

In ReplySetter, which is what I know best:

• Visual job flows: steps with objectives and exits, AI conditions and switches, scenarios for off-script moments. See job flows.
• Full traces of every reply: the step, the reasoning, each tool call. See Test & Debug.
• Your choice of model: Anthropic, OpenAI or 350+ models through OpenRouter, with backup models.
• A/B testing of steps and characters.
• Multi-sub-account agents: one agent serves many sub-accounts, matching calendars, pipelines and workflows by name in each.
• Other channels too: Gmail, Outlook, HubSpot, Instantly, SmartLead and website chat, if a client needs them.

I compare the two in detail on the GoHighLevel Conversation AI alternative page.

Setting up ReplySetter on GoHighLevel, step by step

1. Connect the sub-account with a Private Integration token or Marketplace OAuth on the Accounts page.
2. Pick the channels the agent should answer.
3. Use tag filters so it only answers new leads, or never existing clients.
4. Build the flow: greet → qualify → book, with your objectives.
5. Map objectives to GHL custom fields so answers land in the contact.
6. Pick the calendar(s) and what a booking does: a tag, a pipeline stage.
7. Add a handoff rule that pauses the AI and assigns the conversation to a user.
8. Test in the sandbox, launch in suggest mode, then auto.

There's a full guide on the GoHighLevel AI appointment setter page.

Which should you choose?

• One business, simple FAQs and booking: Conversation AI is a reasonable start.
• You want a real qualification process, testing and visibility: an external setter.
• Agency with many sub-accounts: an external setter that serves many sub-accounts from one agent saves a lot of time.
• You also need email, cold email or HubSpot: an external setter.