Skip to main content

​
Getting Started with Automated Analysis

This guide walks you through creating a basic analysis setup to extract “Next Steps” from your sales calls. We’ll use the recommended Chat Editor for creating and refining our Analysis Field. Goal: Automatically identify and record the agreed-upon next steps after every call tagged as a “Discovery” call.

​
Prerequisites

  • Ensure call recording integration (like Zoom) and CRM integration are set up.
  • Familiarize yourself with the Core Concepts.

​
Steps

  1. Create the Analysis Configuration:
    • Navigate to Settings -> Analysis -> Configurations.
    • Click the “New Analysis” button (or similar action button).
    • In the modal:
      • Name: Discovery Call
      • Type: Standard
      • Mode: Call
    • Click “Create”. You’ll be taken to the configuration details page.
  2. Create the Analysis Field (Using Chat Editor):
    • On the Discovery Call configuration page, scroll down to the “Analysis Fields” section.
    • Click “Create & Link Field”.
    • A modal appears.
      • Name: Next Steps
      • Select “Value” as the field type (since we’re extracting information, not scoring).
      • Purpose: Enter a clear goal, e.g., Extract the specific, agreed-upon action items and next steps discussed during the call, including who is responsible and any deadlines mentioned.
      • Click “Create”. This creates the basic field structure. Crucially, we will now refine the instructions using the Chat Editor.
    • You’ll be redirected to the newly created “Next Steps” field page .
    • Click the “Create Draft” button, and select “Edit Field”. This will take you to the Chat Editor interface for this field’s first draft.
  3. Refine the Field with the Chat Editor:
    • The Chat Editor will open. You’ll see the “Purpose” you entered.
    • The AI assistant will greet you. Start interacting:
      • You: “Based on the purpose, please generate initial instructions for this ‘Next Steps’ field.”
      • (AI generates instructions)
      • You: “Okay, let’s refine that. Ensure the instructions explicitly ask to identify who owns each next step and any due dates mentioned. Format the output as a markdown bulleted list.”
      • (AI proposes updated instructions. You’ll see a diff view (DiffableFieldEditor.tsx)).
      • Review the changes. Click “Accept” if they look good.
    • (Optional) Test the Field:
      • Scroll down to the Test This Field section.
      • Select Interaction Type: Calls.
      • Use the Call Search to find 5-10 recent discovery calls. Select them.
      • Click “Test Field”.
      • Review the “Draft Result” column. Does it accurately capture the next steps from those calls according to your refined instructions? If not, go back to the Chat Editor and iterate further (e.g., “In the last test, you missed the deadline for the first action item. Please adjust the instructions to emphasize capturing deadlines.”). Accept changes and re-test.
    • Once satisfied, click “Deploy Changes” at the top of the Draft Editor page.
  4. (Optional) Add a Filter Template:
    • While still on the configuration page, click “Edit” (if not already editing).
    • Find the “Filter Template” field.
    • Enter a simple filter, e.g., {{ TitleContains "Discovery" }}. This tells Fabius to only run this configuration on calls whose title includes the word “Discovery”.
    • Click “Save”.
  5. Activate the Configuration:
    • Ensure the “Active” switch on the configuration page is turned ON.
    • Click “Save” if you made changes.

​
Viewing Results

From now on, when a call occurs whose title contains “Discovery”, Fabius will:
  1. Match it to the Discovery Call configuration via the filter.
  2. Run the Next Steps analysis field using the instructions you refined and deployed.
  3. The extracted next steps will appear in the Analysis section when you view that specific call’s transcript page.
This simple example demonstrates the core workflow. You can build much more sophisticated analyses by adding more fields (Scores and Values), creating more complex Filter Templates, and leveraging context like Knowledge Documents. Next Steps: