How Manual Review works
To optimize the document verification workflow, Manual Review can be organized into multiple stages handled by different users. Splitting review across stages lets each operator focus on the work that matches their expertise: less experienced operators can clear routine tasks quickly in the first stage, while documents flagged by rule violations or low-confidence extractions are escalated to more experienced reviewers in later stages. This reduces overall processing time and lowers the risk of errors slipping through unchecked. Tenant Administrators can assign the Manual Review Operator role with customized access rights, either to specific skills and stages or to all available skills and stages. Depending on their permissions, an Operator may work across the entire process or only within designated portions of it. At a high level, the manual review process follows these steps:- Task Assignment
The Manual Review Operator opens the application and receives a task containing one or more documents with automatically determined document types and extracted data. - Data Verification
The Operator checks the document type, the recognized data, and the extracted field values. When necessary, the Operator may adjust the document type, correct extracted fields, or modify field regions directly on the document image. - Task Completion or Escalation
After completing the review, the Operator finalizes the task. If additional verification is required — for example, by a more experienced reviewer — the Operator can forward the task to the next operator or stage.
How Manual Review relates to other features
Manual Review sits at the end of the document-processing pipeline, after classification and extraction have run. Several other ABBYY Vantage features shape what Operators see and do in the Client:- Skills define which document types are processed and which fields are extracted, establishing the data that Operators verify.
- Rule verification determines which extractions are flagged for review — rule violations are what typically trigger escalation to a more experienced stage.
- Data catalogs provide lookup values that populate and constrain fields during review.
- Online Learning captures corrections made in the Client and feeds them back to improve classification and extraction accuracy over time.
If a field value is changed during manual review, the data catalog lookup rule runs again. If a matching record is found, the field becomes read-only.
