AI-Guided Configuration

Upload an invoice and see what our AI sees

Upload your sample invoice and explore what Document Intelligence automatically detects:

• Document structure: AI identifies headers, tables, line items, and footer sections
• Field detection: Automatically finds vendor names, dates, amounts, tax information
• Table recognition: Detects line item tables with columns for descriptions, quantities, prices
• Layout understanding: Recognizes document flow and relationships between elements

The AI shows you exactly what it sees with visual highlighting and suggested field names, giving you complete transparency into its understanding of your document.

Create a data model for consistent extraction

Define the structure for how you want data extracted from all future invoices:

• Consistent output: Ensure every processed invoice returns data in the same format
• Standardized fields: Create uniform field names across all document variations
• Reliable structure: Guarantee that agents and downstream systems receive predictable data
• Business alignment: Structure data to match your existing workflows and systems

This data model becomes the template that ensures consistent extraction regardless of invoice format variations.

Select the fields that matter most to you

Choose only the specific fields and tables you need for your business processes:

• Focus on essentials: Select vendor information, totals, dates, and line items that matter
• Custom naming: Rename fields to match your business terminology (e.g., "Supplier" instead of "Vendor")
• Skip unnecessary data: Ignore fields that aren't relevant to your workflows
• Table customization: Choose which columns from line item tables you want to extract

This selective approach ensures clean, focused data extraction that matches your exact requirements.

Define data types and add business context

For each selected field, specify the type of data and provide business context:

• Data types: Define fields as text, numbers, dates, currencies, or boolean values
• Business rules: Explain how fields should be interpreted ("Due date is always 30 days from invoice date")
• Validation requirements: Specify what makes a field valid ("Vendor tax ID must be 9 digits")
• Handling instructions: Guide the AI on edge cases ("If no PO number, use 'N/A'")

This business context ensures the AI understands not just what to extract, but how to interpret and validate the data according to your business logic.

Create your data model

Finalize your configuration to create a reusable data model:

• Save configuration: Your data model becomes the template for processing all similar invoices
• Test with variations: Validate that the model handles different invoice formats correctly
• Refine as needed: Easily adjust field definitions and business rules through the interface
• Deploy for production: Start processing invoices automatically with your configured model

The completed data model ensures consistent, accurate extraction that adapts to document variations while maintaining your specified structure and business rules.