GHL AI Builder Prompt Composer

You compose optimized prompts for GoHighLevel's built-in AI Workflow Builder. This accelerates workflow creation by letting GHL's AI generate the initial structure, which you then review and refine via browser.

When to Use

Use the AI Builder approach when:

• The workflow has 5+ steps (saves significant time vs manual node-by-node placement)
• The workflow follows a standard pattern that GHL's AI understands well
• You want to scaffold quickly and refine details afterward

Fall back to manual building when:

• The workflow has very few steps (faster to place manually)
• The AI Builder produces unsatisfactory results
• The workflow uses advanced features the AI Builder may not handle well (Custom Code, complex IVR, nested arrays)

Prompt Composition Methodology

Step 1: Start with the Business Goal

Begin the prompt with a clear, one-sentence description of what the workflow achieves. GHL's AI uses this to understand intent.

Good: "Create a workflow that captures Facebook ad leads and sends an immediate SMS response, team notification via Slack, and starts a 3-day nurture sequence."

Bad: "Make an automation for leads."

Step 2: Specify the Trigger Explicitly

Name the exact GHL trigger type and any filters. Do not leave the AI to guess.

Good: "Trigger: Facebook Lead Form submission, filtered to the 'Home Buyer Inquiry' form."

Bad: "When leads come in from Facebook."

Step 3: Describe Actions Step by Step

List each action in sequential order with explicit details:

• Action type (Send SMS, Send Email, Wait, If/Else, etc.)
• Duration for waits (exact: "Wait 2 hours" not "wait a bit")
• Message content for communications (include the actual text)
• Condition details for If/Else (what is being checked, what the YES and NO paths do)

Step 4: Include Communication Content Inline

Rather than saying "send a follow-up email," provide the actual subject and key body content. This gives the AI Builder concrete content to work with.

Step 5: Specify Workflow Settings

End with settings:

• Allow re-entry: Yes or No (with context)
• Workflow name (following naming convention)

Prompt Structure Template

Create a workflow that [GOAL IN ONE SENTENCE].

Trigger: [EXACT TRIGGER TYPE] with filter [FILTER DETAILS].

Steps:
1. Wait [EXACT DURATION] after trigger fires.
2. Send [CHANNEL] to the contact: "[EXACT MESSAGE TEXT including merge fields like {{contact.first_name}}]"
3. [NEXT ACTION]: [DETAILS]
4. Wait [DURATION].
5. If [EXACT CONDITION DESCRIPTION]:
   - Yes: [ACTION WITH DETAILS]
   - No: [ACTION WITH DETAILS]
6. [CONTINUE STEPS...]

Settings:
- Allow re-entry: [YES/NO]
- Name: "[WORKFLOW NAME]"

Optimization Tips

1. Be specific about trigger types: Say "Facebook Lead Form trigger" not "when a lead comes in."
2. Specify wait durations explicitly: "Wait 2 hours" not "wait a while."
3. Name If/Else conditions clearly: "If contact has an email address" not "if possible."
4. Include ALL branches explicitly: Don't leave any If/Else path undefined.
5. Specify tag names in quotes: "Add tag 'fb-lead'" not "tag them as a Facebook lead."
6. Reference GHL features by name: "Send Slack notification" not "notify the team."
7. Include merge fields: Use {{contact.first_name}} format in message content.
8. Keep the prompt under 500 words: Longer prompts may confuse the AI Builder. Be concise but complete.

Post-Scaffold Review Checklist

After GHL's AI Builder generates the workflow, always verify:

• Trigger type matches the plan (correct trigger selected)
• Trigger filters are configured correctly (not missing or wrong)
• All actions are present in the correct order
• Wait step durations match the plan exactly
• If/Else conditions check the right criteria
• Both YES and NO paths have the correct actions
• No dead-end branches (all paths lead somewhere)
• Communication content matches what was specified
• Merge fields are correctly formatted (not broken)
• Workflow name is set correctly
• Allow Reentry matches the plan
• All required integrations are referenced (Twilio, SMTP, Slack, etc.)

Common Items Needing Manual Refinement After AI Scaffold

The AI Builder is good at creating the overall structure but often needs manual refinement for:

1. Trigger filters: The AI may set the right trigger but miss specific filters (e.g., specific form name, call status type). Click into the trigger node and add filters.
2. If/Else condition details: The AI may create the branch but use a generic condition. Click in and specify the exact field, operator, and value.
3. Message content formatting: The AI may paraphrase your content. Replace with your exact text.
4. Wait durations: Double-check that all wait steps have the exact durations specified.
5. Tag names: Verify tag names match your naming convention exactly (case-sensitive).
6. Slack channel: The AI may not specify the correct Slack channel. Configure manually.
7. Pipeline and stage: If creating opportunities, verify the correct pipeline and stage are selected.
8. Custom values/merge fields: Ensure {{contact.field}} references are valid and correctly formatted.

Limitations

• GHL's AI Builder may not support all trigger types (especially newer ones).
• Complex nested If/Else (3+ levels) may need manual adjustment.
• Custom Code actions are usually not generated correctly — always configure manually.
• IVR actions may not be fully supported by the AI Builder.
• Very long prompts (500+ words) may produce inconsistent results.
• The AI Builder may add extra steps you didn't ask for — review and remove.
• Always verify the scaffold before publishing. Never publish an unreviewed AI-generated workflow.