Skip to main content

​
Configuring Value Fields

Value Fields are used when you want the AI to extract specific pieces of information from an interaction or generate new text content based on it.

​
When to Use Value Fields

  • Extracting structured data (e.g., Next Steps, Competitors Mentioned, Stakeholders Identified).
  • Summarizing key topics or sections of a call/document.
  • Generating content (e.g., Follow-up Email Drafts, Call Summaries).
  • Identifying specific mentions (e.g., Budget Discussed, Pain Points Voiced).

​
Creating a Value Field

  1. Follow the steps in Managing Fields.
  2. During creation, select “Value” as the field type.
  3. Define a clear Purpose (description) explaining what information this field should capture or generate and how it will be used.
  4. Proceed to the Chat Editor to define the core prompt components.

​
Defining Instructions

For Value Fields, the Instructions section is paramount. It tells the AI exactly what to do.
  • Clarity is Key: Be explicit about the information to extract or the content to generate.
  • Specify Format: Instruct the AI on the desired output structure. Even though the underlying type is often treated as a string in the UI, you can request specific formatting within that string:
    • “Output as a Markdown bulleted list.”
    • “Generate a concise, 3-sentence summary.”
    • “List each identified stakeholder on a new line.”
    • “Format the output as ‘Key: Value’ pairs, one per line.”
  • Handling Absence: Tell the AI what to output if the requested information isn’t found (e.g., “If no competitors are mentioned, output ‘None mentioned’.”).
  • Break Down Complexity: For complex extractions, list numbered steps in the instructions.
Example Instructions (Next Steps Value Field): 
1.  Analyze the call transcript to identify all specific action items or next steps agreed upon by participants.
2.  For each next step, determine:
    - The action itself.
    - Who is responsible for completing the action (e.g., "Rep", "Prospect", "John Doe"). If unclear, state "Responsibility Unclear".
    - Any deadline or timeframe mentioned. If none, state "No Deadline Mentioned".
3.  Format the output as a Markdown bulleted list. Each main bullet should be the action item. Use sub-bullets for "Owner:" and "Due:".
4.  If no next steps were clearly agreed upon, output the single phrase "No clear next steps identified."

Example Output Format:

- Schedule follow-up technical deep dive
  - Owner: Rep
  - Due: Next Tuesday
- Prospect team to review security documentation
  - Owner: Prospect (Jane Smith)
  - Due: End of Week

​
Using Directions for Values

Use the Directions section to:
  • Provide examples of desired vs. undesired output.
  • Clarify how to handle ambiguous mentions.
  • Specify the desired level of detail or tone (e.g., for generated emails).
  • Give guidance on extracting information that might be phrased indirectly.

​
Advanced Value Field Options

While editing a Draft (via manual or Chat Editor), you might encounter these advanced options:
  • Custom Output Message: Provides a template around the AI’s generated output, often used for structuring complex outputs or adding standard framing text. (Less common for simple string extractions).
  • Scratchpad: Instructs the AI to perform step-by-step reasoning before generating the final output. Useful for complex tasks where intermediate thinking helps accuracy (e.g., calculating a potential ROI before summarizing it). Enter the prompt for the scratchpad thinking process here.
Value Fields are versatile tools for turning unstructured conversation data into structured information or automated content. Clear instructions are the key to their success. Next: Understanding Field Dependencies