eye-share Capture FAQs

AI Capture is an AI-powered document interpretation engine in eye-share Capture that automatically extracts and understands information from documents such as invoices.

Unlike traditional OCR, AI Capture analyzes both the content and context of a document to identify relevant information.

Traditional OCR primarily focuses on reading text from predefined areas or templates.

AI Capture goes further by:

• Understanding document structure and meaning

• Adapting to different invoice layouts

• Handling variations in formatting

• Interpreting complex scenarios and poor-quality documents

• Supporting handwritten and machine-written content

• Using AI reasoning rather than fixed extraction rules

AI Capture is designed to process a wide range of business documents, and the AI engine is not dependent on specific layouts or setups to extract the correct information.

AI Capture does not rely on predefined templates, zones, or patterns in the same way as traditional OCR solutions. Instead, it uses prompts and contextual understanding to identify and extract information from documents.

Patterns are not part of Capture AI. Capture AI uses artificial intelligence to identify and extract information from documents based on prompts, context, document structure, and content.

For example, the Invoice Number field is based on a generic prompt that works across all documents in a folder or job. There are no supplier-specific patterns or configurations, which means the maintenance traditionally required for individual suppliers is no longer needed.

The pattern jobs/profiles can be stored so customers can return from Capture AI to Capture with pattern without creating setup all over again.

A prompt is a field-specific description that helps Capture AI understand what information should be extracted from a document. Instead of relying on supplier-specific rules or patterns, the prompt describes the type of information the AI should look for, such as an invoice number, invoice date, PO number, total amount etc.

Capture AI uses the prompt together with the document's content, structure, and context to identify the most relevant value for a field. The prompt acts as guidance for the AI, helping it distinguish between similar values that may appear on the document.

Prompts can be refined to better describe the information being extracted. Clear and well-defined prompts can improve extraction accuracy, especially fields that may be represented differently across documents. Prompts can only be edited by superusers.

AI Capture offers extensive language support and can process documents in more than 300+ printed languages and subset of handwritten languages.

• Low-quality scans
• Difficult invoice layouts
• Handwritten content
• Mixed handwritten and printed documents

• At profile level
• At individual field level

• Supplier
• Invoice number
• Invoice date
• Purchase order number
• Total amount

AI Capture provides field-level traceability showing the source of extracted values and confidence levels, helping users understand how information was identified and why it was selected.

• The document may contain "US Dollar"
• AI Capture returns "USD"

• The document must be verified by a user before it can be sent from capture. Either by accepting the information found or by changing to a different value
• Business expressions can be applied as an additional validation layer

AI Capture uses the same familiar eye-share Capture interface to keep training requirements at a minimum. The primary change is the interpretation engine running behind the scenes. Training to introduce setup, use etc will be provided as a part of delivery.