Extractions

Extraction Tools: Pulling Key Data from Conversations

Extraction tools help your AI assistant capture specific pieces of information from a conversation—like a client’s budget, preferences, or other custom details. Here’s how to set them up, when to use them, and best practices for reliability.

How to Set Up an Extraction Tool

Open Tools & Abilities:
In your AI editor, click “Manage Tools” and select the Extractions tab at the top.
Create a New Extraction Tool:
- Click “Create New Tool.”
- Give it a unique name (e.g., extract_budget).
- In the description, specify when and why to use it (e.g., “Use this tool to fill the user's home budget every time they share it.”).
Map to a Custom Field:
- Click “Choose Custom Field” and select the field you want the data mapped to (must be created in your CRM before this step).
Describe the Value:
- Add a value description (e.g., “Fill this with the user's home budget”).
Save and Add to Assistant:
- Click “Save Changes,” then “Add to Assistant” to activate the tool.

Prompt Integration for Better Consistency

Mention the Tool Name in Your Prompt:
For example:
“When the user shares their budget, use tool extract_budget to save the user's budget.”
Place the Mention Near the Relevant Prompt Section:
This ensures the AI knows when to trigger the tool.

💡 Tip: Always align your prompt language with your tool names (using underscores, all lowercase) for best results.

When to Use (and Not Use) Extractions

Live Extractions:
Use when you need information pulled immediately during the conversation (e.g., booking flows, qualifying questions).
Automatic Extractions (Recommended):
For most use cases, post-call/chat extractions are simpler and add less latency. See your “automatic extractions” tools for this setup.
Complex Data or Multiple Choices:
If you need to extract from a set list (like colors or product types), you can set up extraction tools for each option. However, in complex cases, using a tag tool or automatic extraction may be more reliable.