1. Overview
  2. How To
  3. How to Extract Important User Information using automatic data extraction?

How to Extract Important User Information using automatic data extraction?

Purpose: To configure Assistable.ai to automatically extract specific information (e.g., Name, Email, Budget) from voice and chat conversations and sync it to GoHighLevel (GHL) Custom Fields with zero latency and 100% reliability.

Scope: This procedure covers the "Automatic Data Extraction" method, which is the recommended approach over using "Extraction Tools" (live function calling) due to its superior reliability and speed.

Prerequisites:

1. GoHighLevel Access: You must have a GHL Sub-account with the target Custom Fields already created.

2. Assistable.ai Access: You must have an Assistant created and ready for configuration.

Phase 1: Preparation (GoHighLevel)

Before configuring the AI, ensure the destination exists.

  1. Navigate to Settings: Open your GHL Sub-account and go to Settings > Custom Fields.

  2. Create the Field: Click Add Field.

    • Recommended Type: Select "Single Line Text" (String).

    • Why? Text fields are the most forgiving. The AI can dump raw data (like "$5k-10k") into a text field safely. .

  3. Note the Name: Copy the exact name of the field (e.g., client_budget).

image.png

Phase 2: Assistant Configuration (Assistable.ai)

Step 1: Access Extraction Settings

  1. Open your Assistable.ai Agency Dashboard.

  2. Click on the Assistants tab and select the assistant you are configuring.

  3. On the right-hand "Global Settings" menu, click Map Custom Fields.

    • Visual Reference: See the red box in the image below.

image.png

Step 2: Create New Extraction

  1. Click the "New Data Extraction" button (or the + icon).

  2. A modal window labeled "Data Extraction" will appear.

image.png

Step 3: Configure Data Points

Fill out the fields in the modal exactly as follows to ensure the "Observation Model" works correctly.

  • Name: * Enter a clear, internal name for this data point (e.g., Customer Name).

  • Type (Data Output Type): * Select string.

    • Note: Even for numbers, "string" is often safer to capture nuances like "around 500".

  • Description (What data to extract): * CRITICAL: This is the prompt for the Observation Brain. Be descriptive.

    • Bad: "Budget"

    • World-Class: "The maximum monthly budget the user is willing to spend. If they mention a range, extract the higher number."

  • Custom Field Mapping: * Click the dropdown and select the GoHighLevel Custom Field you created in Phase 1.

image.png

Step 4: Save and Verify

  1. Click Save Changes.

  2. The system will now automatically listen for this data.

    • Voice Calls: Data extracts Post-Call (immediately after hang-up).

    • Chat: Data extracts Per-Run (updates after every message exchange).

Quality Assurance & Best Practices

Do

Don't

Do use detailed descriptions in the "Description" field. Treat it like a mini-prompt.

Don't use "Extraction Tools" for simple data entry. It pauses the conversation to "think."

Do use "String" (Text) fields in GHL to prevent validation errors.

Don't try to force complex logic (like "If X then Y") inside the extraction. Do that in GHL workflows.

Do test by having a natural conversation. The AI should not "ask" for the field if it already heard it naturally.

Don't worry if the main AI doesn't acknowledge the extraction immediately; the Observation Model works silently.

Troubleshooting:

  • Data not showing up? Check that your "Description" explicitly tells the AI what to look for and that the "Custom Field Mapping" is connected to the correct GHL field.

  • Latency issues? Ensure you are strictly using Map Custom Fields (Automatic) and not the "Tools & APIs" section for this task.

Was this article helpful?