Data Fields

Configure extraction schemas for your desk

Data Fields

Data fields define what information Kantilever extracts from submission documents. Configure fields to capture the data most relevant to your underwriting workflow.

Field Types

Type	Description	Example
Text	Free-form text	Insured name, address
Number	Numeric values	Building count, years in business
Currency	Monetary amounts with currency	Total insured value, premium
Date	Calendar dates	Policy effective date, expiration
Boolean	Yes/No values	Prior losses?, Claims-made?
Enum	Predefined options	State, construction type
Object	Structured data	Location details
Array	Lists of values	Multiple locations, coverage limits

Creating Fields

1. Navigate to your desk’s Data Fields tab
2. Click + Add Field
3. Configure the field:
   • Name - Identifier used in rules and API
   • Display Name - Human-readable label
   • Type - Select from available types
   • Required - Whether this field must be extracted
   • Description - Help text for extraction AI

Enum Fields

For enum fields, define the allowed values:

States: AL, AK, AZ, AR, CA, CO, CT, ...
Construction: Frame, Joisted Masonry, Non-Combustible, Fire Resistive

Extraction Behavior

When a submission is created, Kantilever:

1. Processes all documents (email body, attachments)
2. Extracts text from PDFs and images
3. Uses AI to identify field values based on your schema
4. Populates fields with extracted values and confidence scores

Confidence Scores

Each extracted value includes a confidence score (0-100%). Review low-confidence extractions for accuracy.

Custom Prompts

Add custom extraction prompts to guide the AI:

“For TIV, sum all building values and contents values. Include business income if specified.”

Computed Fields

Create fields that calculate values from other fields:

1. Click + Add Computed Field
2. Define the computation logic
3. Reference other fields in your formula

Example: Total Premium = Base Premium * (1 + Tax Rate)

Best Practices

Start Simple

Begin with essential fields and add more as needed:

• Insured Name
• Effective Date
• Expiration Date
• Total Insured Value
• State/Location

Use Clear Names

Field names should be unambiguous:

• Good: policy_effective_date
• Avoid: date1

Provide Context

Add descriptions to help the AI understand what to extract:

• “The named insured, not the broker or agent name”
• “Total property value including all locations”

Review and Refine

Monitor extraction accuracy and adjust field definitions as needed.