Skip to main content
Imagine your WhatsApp replying to customers, answering their questions, and booking appointments on its own. With Conversation AI Agents and your GoGHL.ai channels, that’s exactly what you get:
  • A customer messages your business on WhatsApp
  • The AI agent reads the message, understands the intent, and replies instantly
  • If they want to book an appointment, it offers available slots from your calendar
  • If they have a question, it answers from your knowledge base
  • If it doesn’t know the answer, it hands off to a human
All automatic, all on WhatsApp, from your own number connected via GoGHL.ai.

​
Prerequisites

Before getting started, ensure you have:
  • An active GoGHL.ai account with the marketplace app installed on your sub-account
  • WhatsApp connected to the sub-account (see WhatsApp Setup)
  • Conversation AI enabled on your CRM plan
  • A calendar configured in your CRM (required for appointment booking)

​
Step 1: Open Conversation AI

1

Select Your Sub-Account

Log in to your CRM and switch to the sub-account you connected to WhatsApp via GoGHL.ai.
2

Go to AI Agents

In the left sidebar, click AI Agents, then open the Conversation AI tab.
Conversation AI Agents dashboard
The dashboard shows how your agents are performing: unique contacts handled, actions triggered, appointments booked, and time saved.

​
Step 2: Create a Bot

1

Click Create Bot

Click the + Create Bot button in the top-right corner.
2

Choose the Bot Type

Pick how you want to build your bot:
  • Prompt Based Bot: provide custom instructions (a prompt) for lead capturing, appointment booking, or general Q&A. Recommended if you are comfortable writing prompts.
  • Flow Based Builder: a visual flow builder for advanced lead capturing and nurturing with complex logic and branching. Recommended for advanced users.
Choose between Prompt Based Bot and Flow Based Builder
This guide uses the Prompt Based Bot.
3

Pick a Template

Select a starting template and click Continue:
  • General Q&A: answers common questions instantly; perfect for FAQs and quick resolutions
  • Appointment booking: helps customers select available slots and confirm appointments
  • Marketplace Templates: pre-built ConversationAI templates (agency admins only)
  • Start from Scratch: build a fully custom bot
Pick a bot template
This guide follows the Appointment booking template, the most common use case for WhatsApp automation.

​
Step 3: Configure Bot Settings

The bot editor is split into two tabs at the top: Build, where you set up how the agent behaves, and Deploy, where you choose the providers it runs on.
Build tab with model, business name, prompt, and settings panels
1

Name Your Bot

Click the pencil icon next to the bot name at the top of the page and give it a recognizable name, for example WhatsApp Booking Assistant.
2

Choose the Model and Business Name

In the Build tab, pick the Model that powers the bot (for example, OpenAI GPT 4.1). Optionally set a Business Name. The bot uses it when referring to your business in replies.
3

Set the Bot Mode

Expand Mode in the right-hand panel and choose how the bot operates:
  • Off: Conversations AI is turned off
  • Suggestive: the bot suggests replies inside the chat window, and you decide what to send
  • Auto Pilot: the bot replies automatically based on its trained data
Start with Suggestive while testing, then switch to Auto Pilot when you’re confident in the bot’s replies.
4

Open the Deploy Tab

Switch to the Deploy tab. Each channel (SMS, WhatsApp, Instagram, Facebook, TikTok, Live chat, Chat widget, and Email) has its own card. Click Add channels from marketplace.
Deploy tab with a card per channel
GoGHL.ai is a marketplace app, so its providers (WhatsApp QR, iMessage, and SMS Android) are added through Add channels from marketplace.
5

Select the Provider

In the Add channel dialog, turn on Assign channel. Open the Installed providers dropdown and select the GoGHL.ai provider the bot should reply on, for example WhatsApp QR. Only providers installed from the marketplace appear in this list.
Add channel dialog with WhatsApp QR selected as the installed provider
6

Filter by Tags (Optional)

Leave the tag fields blank to run the bot on every conversation for that provider. To narrow which conversations it handles, use Has tags and Doesn’t have tags. Use the AND / OR toggle to decide whether a contact must match all of the tags or any one of them.
Has tags filter with a tag selected
7

Add the Channel and Save

Click Add channel, then click Save at the top right.

​
Deploy Different Agents for Different Providers

Each agent is deployed to a specific provider, so you can run a separate Conversation AI agent for each one:
  • One agent per provider. For example, deploy a booking agent on WhatsApp QR and a support agent on iMessage or SMS Android.
  • One agent per number. To make a bot respond only for one number, add that number’s tag under Has tags. The bot then triggers only for conversations that carry that tag.
  • No workflow needed. Provider and tag targeting live in the agent’s Deploy settings instead of in workflow logic.

​
Step 4: Configure Timing and Response Behavior

Fine-tune how the bot behaves when replying automatically. These settings live in the collapsible panels on the right side of the Build tab.
Mode and Timing & Pacing panels
Timing & Pacing
  • Wait time before responding: how long the bot waits before replying (seconds, minutes, or hours). A short delay lets customers finish typing multi-part messages.
  • Max messages per conversation: caps bot replies per conversation to control AI usage and costs.
  • Send bot to sleep when I send a: pause the bot when a Manual Message or Workflow Message goes out. Enable both to pause on any outbound message, so the bot never talks over you or your existing workflows.
Response Behavior and Summary Settings panels
Response Behavior
  • Enable Response Style Settings: choose how the bot writes: concise, balanced, or detailed.
  • Also allow this bot to respond to: enable Images and/or Voice Notes so the bot can understand them. These responses may take slightly longer and are supported only in Auto-Pilot mode.
Summary Settings
  • Enable Conversation Summary: generate an automatic summary when a conversation goes quiet, and send it to a specific user or a custom email.
The Mode panel also has Configure Email Settings for the Email channel and a Workflow & Transfer bot channels link for bots triggered from workflows or other bots.
Click Save when done.

​
Step 5: Train the Bot

The bot answers from your knowledge base. Connect one from the Knowledge Base Triggers panel in the Build tab.
Adding a knowledge base trigger
1

Create a Knowledge Base

If you don’t have one yet, click Create New under Add Knowledge Base Trigger, give it a name and an optional description, and click Create.
2

Add Content

Add sources the bot can learn from:
  • Website: import an exact URL or all URLs in a domain; pages are fetched and indexed automatically
  • FAQs: add question and answer pairs
  • Tables: structured data such as price lists
  • Rich text: free-form content
  • Upload file: documents like PDFs
3

Add a Knowledge Base Trigger

Expand Knowledge Base Triggers and click Add Knowledge Base Trigger. In the Add Knowledge Base dialog, choose your knowledge base under Select knowledge base (you can select up to 7), and optionally add instructions such as:
Click Save.
Instructions are optional; the agent autonomously decides when to use each knowledge base. Adding instructions makes routing more predictable.

​
Step 6: Write the Bot’s Prompt

The large prompt editor in the Build tab defines how the bot talks and what it should achieve. It starts with a template split into three sections: Personality, Goal, and Instructions.
Prompt editor with Personality, Goal, and Instructions sections
  • Personality: describe the tone and role, for example:
  • Goal: what the bot should drive toward. With the Appointment booking template, the goal is to get customers to book an appointment; you can extend it, for example to answer questions about features, pricing, setup, and integrations.
  • Instructions: conversation guidelines, examples, and rules, for example:
Click Custom Values to insert dynamic values such as your business name. The counter above the editor shows the approximate token count and how many of the 2,000 words you have left. For more tips, open Prompt Guidelines below the editor.

​
Step 7: Set Up Actions

Actions let the bot do more than chat. In the Build tab, expand the Actions panel and click Setup Your Actions.
Connect a single calendar or multiple calendars and click Proceed. Then configure:
  • Book via link: send customers a booking link instead of booking in-chat (leave off to let the AI book directly)
  • Pause bot response after booking: stop the bot once an appointment is booked
  • Trigger workflows after booking: fire a workflow for confirmations or reminders
  • Transfer to employee post appointment booking: assign the conversation to a specific team member
  • Allow the bot to cancel appointments and reschedule appointments: enable both so customers can manage bookings entirely in chat
Let the bot collect information during the conversation (for example, city) and automatically save it to the contact record. You can add new fields directly from this screen.
When a customer says “bye”, “goodbye”, or “thank you, have a nice day”, the bot stops responding. This also prevents wasted AI usage. Optionally set:
  • A final message (e.g. “Thank you for your time, have a nice day!”)
  • Reactivate bot after a set number of minutes, hours, or days, or keep it off
  • A custom tag so you can track conversations where the bot stopped
Hand the conversation to a real person in three scenarios:
  1. Human requested: triggers on phrases like “I want to talk to a manager” or “I want to talk to a human”. Assign the conversation to a specific user, and enable Skip assigning if contact has an assigned user to avoid reassigning.
  2. Can’t find relevant information: the bot lacks an answer. Set a final message such as “I’m sorry, I don’t have the answer to your question at the moment. I’m forwarding your request to our team and someone will get back to you within 24 hours.”
  3. Failed to resolve issue: multiple attempts to resolve the issue have failed; the conversation is escalated to your team.
Pair human handover with GoGHL.ai internal notifications to get pinged on WhatsApp or email the moment the AI can’t answer, and jump in to reply yourself.
Route the conversation to another bot: either a default transfer to your primary bot when this bot doesn’t know the answer, or scenario-based transfers to specialized bots you’ve created.
If a contact stops replying, the bot follows up automatically after the delay you set (minutes, hours, or days), with multiple follow-up steps if needed. If the contact says they’re busy (“contact me after 15 minutes”), the bot honors the requested time.

​
Step 8: Test Your Bot

Use the Test your agent panel on the right side of the Build tab before going live. Save the bot first, then send a message the way a lead would. Test messages are never sent to your contacts.
  • Ask questions your customers would ask: pricing, features, booking a demo
  • Try booking: the bot offers real slots from the calendar you connected
  • Rate answers with the like/dislike buttons to flag wrong responses
  • Open a reply’s details to see the prompt and intent used, chat history, and actions executed
When answers look right, set the Mode to Auto Pilot in the Build tab, click Save, and your WhatsApp AI agent is live.

​
Best Practices

  • Start in Suggestive mode and review the bot’s suggestions before enabling Auto Pilot
  • Keep your knowledge base focused: connect separate knowledge bases for pricing, support, and product info with clear trigger instructions
  • Set a sensible maximum messages cap to control AI usage
  • Enable goodbye detection to avoid unnecessary replies
  • Always configure human handover so customers are never stuck with a bot that can’t help
  • Enable send bot to sleep on manual message so the bot never interrupts your team’s conversations

​
Troubleshooting

Bot doesn’t reply on WhatsApp
  • Make sure you have money in your wallet (check View credits in the top bar). Without balance, Conversation AI will not work.
  • In the Deploy tab, confirm you added a channel through Add channels from marketplace with your GoGHL.ai provider (for example WhatsApp QR) selected under Installed providers
  • Check the bot status is Auto Pilot (in Suggestive mode it only suggests, never sends)
  • Verify your WhatsApp number is connected in GoGHL.ai (see WhatsApp Setup)
Bot gives wrong or empty answers
  • Check the knowledge base finished indexing your website pages
  • Add trigger instructions so the right knowledge base is used
  • Tighten the Personality and Instructions sections of the prompt
Appointments not booking
  • Ensure a calendar is connected in the Appointment Booking action
  • Confirm the calendar has available slots

​
Support

If you need help setting up your Conversation AI agent:

​
FAQ

Yes. The bot replies through the WhatsApp number you connected to your sub-account via GoGHL.ai, on any channel you enable.
In the Deploy tab, click Add channels from marketplace and select your GoGHL.ai provider, for example WhatsApp QR, under Installed providers. GoGHL.ai is a marketplace app, so its providers (WhatsApp QR, iMessage, and SMS Android) are listed there.
Yes. Create multiple bots (for example, one for Q&A and one for booking) and deploy each one to a different provider, such as WhatsApp QR or iMessage, from its Deploy tab. You can also use the Transfer Bot action to route conversations between bots, and set one bot as the Primary Bot.
Not if you enable Send bot to sleep when I send a Manual Message in the Timing & Pacing panel. The bot pauses whenever you step into the conversation.
Yes. Enable Images and Voice Notes in the Response Behavior panel. These replies may take slightly longer and are supported only in Auto-Pilot mode.
You’re now ready to run a fully automated WhatsApp AI agent using GoGHL.ai + Conversation AI.