Skip to main content
Documentation
Documentation
Last Updated: 2025-11-14
Configure a Custom AI Model

Configure a Custom AI Model

  • Enable Custom AI in your workspace.
  • Identify the field you want to automate.
  • Gather the set of documents for analysis.
  • Identify the data point you want to track for each document.
Configure a custom AI field model to automatically extract and populate data from business documents into a custom field. Contract Management and Document Intelligence uses natural language instructions to identify and extract the specified information.
You can also use the Custom AI Model Library to save time and effort by eliminating the need to manually build your own data fields. If you don't see what you need in the Model Library, you can still configure your own Custom AI model. See Add Models from the Custom AI Model Library
  1. Click
    Automation
    .
  2. On the
    Fields
     tab, click
    New Field Model
    .
  3. Click
    Get Started
    .
  4. Select a custom field.
    The AI model populates this field automatically.
    Contract Management and Document Intelligence
     extracts data in these formats based on the field type:
    Field Type
    Use When
    Date
    You want to extract specific dates.
    Number
    You want to extract numerical values for reporting and calculations.
    Text
    You want to extract text in various formats.
    Drop-down
    You want to categorize the extracted information into a predefined list of values.
  5. Define the set of documents for the model to analyze.
    You can apply these filters to documents:
    • Clause Type (presence or absence)
    • Contract Management and Document Intelligence Standard AI Fields
    • Contract Type
    • Effective Date
    • Folders
    • Upload Date
    You can improve model accuracy and manage processing costs by applying filters. Contract Management and Document Intelligence runs the AI model only on documents that match all filter criteria. Contract Management and Document Intelligence processes the entire document, even when you scope it to a specific clause type.
  6. Enter natural language text that describes the data you want to extract in the instructions field.
    You can use these examples of effective instructions:
    • What to extract: The specific data point or answer you need. Example: Extract the total contract value.
    • How to find it: The logic to use to identify the correct value. Example: Sum the values from the annual fee schedule. Exclude one-time implementation fees.
    • Output format: The desired format for the answer, especially for text or drop-down fields. Example: Return Yes if the customer can use our logo. Return No if they can't. Return Not Mentioned if the topic isn't referenced.
    • Context: Your company's name or role in the document to clarify obligations. Example: Identify if we, Global Modern Services Inc., have the right to assign the agreement.
    • Edge cases: How to handle ambiguity or conflicting information. Example: If the agreement spans multiple years with different payment amounts, extract the value for the first year only.
  7. Select
    Save and Evaluate
    .
  8. Review each document and rate the model's result.
    Select these values:
    • Correct: The model returned the correct value.
    • Incorrect: The model returned an incorrect value. Enter the correct answer when prompted. You can optionally provide an explanation.
    • Skip: You want to ignore the document during evaluation.
    If you receive incorrect results, update your instructions to provide more clarity or address ambiguity. Select
    Save and Evaluate
     again to test your changes against the same set of documents.
    Repeat the evaluation process until the model consistently returns accurate results. We recommend you review 10 to 25 documents.
  9. Select
    Review & Publish
    .
    Finalize and activate the model.
Workday creates and deploys the custom AI model. The model runs on all existing documents within the defined scope and on all new documents that meet the scope criteria. The model populates the selected custom field with the extracted data.
Use the
Custom AI Models
 report to monitor model accuracy and usage. You can edit the model to refine instructions or change the scope at any time.