Within the Affinda platform, rules can be configured to validate automatically if the data extracted by our AI models meets certain business logic, and therefore does not need any user intervention or review. Ultimately, this means that straight-through processing of documents can be achieved to remove the 'human in the loop' from the process. The rules can also be used to flag exceptions where the data does not meet expected values, ensuring accurate data for downstream processing.

These rules are surfaced to users in the validation UI, showing users both which rules have passed (and therefore which fields have been auto-validated) and any failed rules that need user attention.

📘

Validation rules are configured by Affinda team

If you would like to configure validation rules in your environment, please speak to the Affinda team who can discuss the options available in more detail and what would work best for your use case.

Types of operations

Some of the standard operations that the Validation Rule engine supports are:

• Regex - does the data extracted meet a defined regex pattern (e.g. IBAN number = [a-zA-Z]{2}[0-9]{2}[a-zA-Z0-9]{4}[0-9]{7}([a-zA-Z0-9]?){0,16})
• Equality Operation - does one field equal another field or value (e.g. does invoice total = invoice due or is there some other element like a partial payment)
• Greater Than / Less Than - check if one field is >/< another or value (e.g. due date > invoice date, invoice date < today)
• Sum - check if the sum of multiple fields / values equals another field / value (e.g. line item totals = document total, subtotal + tax = total)
• Length - check that number of characters in field is within specified range (e.g. is PO number 8 digits?)
• ABN Validation - confirms that ABN is valid

If you have other rules you would like to setup, please speak to the Affinda team.