This guide walks you through how to extract important information from conversations using tools and sync that data into custom fields in Agent Suite. We'll cover how to create and connect extractions, how to prompt for them, and where that data ends up.
Go to the right-hand sidebar and click Tools and Abilities.
Select Manage Tools → scroll to Extraction.
Click Create a Tool and choose Extraction as the type.
Name your tool (e.g., extract_user_business_name).
Use lowercase and underscores — avoid spaces. It must follow JSON formatting.
Paste in the extraction description (example below).
Example Description:
“Always use this tool to extract and store the user's business name so a team member can reference the information at a later date.”
You must create custom fields in Agent Suite before syncing tools.
Once created, search your field (e.g., “Business Name”) when prompted in the tool setup.
Click to sync it with the extraction tool.
You can create similar tools for:
Business Description
Best Selling Service
Session Feedback
Use the same structure and formatting for each. Add them all to your assistant after setup.
It’s not required to prompt every extraction, but doing so improves consistency. Here’s how you should prompt it inside your task list:
Step 5 (Business Name):
“Use the tool extract_user_business_name to extract and store the user's business name. This is an internal tool, so don’t mention it to the user. Continue to step 6 after calling the tool.”
Step 7 (Business Description):
“Call the tool extract_user_business_description to extract and store the user's description. Do not use the tool output in your response. Move to step 8 afterward.”
Step 9 (Best Selling Service):
“Call the tool extract_user_service to extract and store the user's best-selling service. Do not include tool responses in chat. Proceed to step 10.”
Step 13 (Session Feedback):
“Call the tool extract_user_feedback to extract and store session feedback from step 12. This is internal. Move to step 14 afterward.”
These tools are invisible to the user unless you include their output in the assistant’s response.
Use phrasing like “move to step X after calling the tool” to maintain flow and keep the tool call silent.
If you're using external tools (e.g., via Make, Buildship), you can return dynamic values and display them if desired.
Once synced, all extracted data will be stored in the mapped custom field.
You can view this data inside the contact record in Agent Suite.
Custom fields used in the demo: Business Name, Business Description, Best Selling Service, Session Feedback.
Create and name your custom fields before setting up extraction tools.
Use internal phrasing to keep tool calls hidden from the user.
Prompting the assistant to use tools increases accuracy.
Always add tools to the assistant after setup — otherwise they won’t be used.
