Quality Control (QA/QC) processing pipeline that Maestro runs when a loan record enters the system, from document ingestion through classification, extraction, and business rule evaluation to reviewer decisions.

This page describes the end-to-end processing pipeline that QA/QC runs when a loan record enters the system, from document ingestion through reviewer decision-making.

The QA/QC workflow is orchestrated by UiPath Maestro. Processing begins when a loan record is submitted to the Maestro process, either automatically by the customer's dispatcher RPA or manually through the QA/QC application. Maestro then handles the full processing sequence: from retrieving documents and classifying them to extracting their contents, and writing structured findings to Data Fabric.

Reviewers then work through the findings in the QA/QC application.

A loan record enters the processing pipeline through the Maestro process trigger. The request carries three fields: Version (the contract schema version), SourceLoanId (the unique loan identifier), and SourceLoanType (the loan type, used to select which business rules apply).

In the Simple implementation, the trigger fires when a reviewer submits the Add Loan form in the application. In the Preferred implementation, the trigger fires automatically when the customer's dispatcher RPA detects a new loan record in the source system and calls the Maestro process's API endpoint.

The Performer retrieves the loan's document package from the institution's document storage system. How the documents arrive depends on the implementation:

Documents may include any combination of the expected document types for the loan product. The system does not require a complete package at this stage; missing documents are flagged later in the classification step.

IXP processes each document in two steps:

Documents that cannot be classified are counted as Unclassified in the Loan documents panel. Required document types that are absent from the package are counted under Missing docs in the loan's QC status.

After extraction, the QA/QC agent evaluates each document against the business rules configured for the loan product. Business rules are stored in Data Fabric and are configured with the institution during implementation.

The Performer writes the complete processing output to UiPath Data Fabric: the loan record with updated status, document metadata, extracted field values, and business rule findings. From this point forward, the QA/QC application reads all data from Data Fabric; the Performer's role in the workflow is complete.

The loan status transitions based on the processing and review outcome. The available statuses are:

After agent processing completes, a reviewer opens the loan in the QA/QC application and begins working through the document checklist. For each document, the reviewer opens the document review view, which presents the business rule checklist alongside the document image and the extracted field values.

For each item, the reviewer examines the evidence: the extracted value, the agent's reasoning, and any cross-referenced value, and records a Human evaluation of Yes, No, or N/A. The decision is saved immediately, recorded with a timestamp and the reviewer's identity in the activity timeline, and the rule collapses to a summary showing the agent and reviewer evaluations.

After the reviewer has evaluated all documents, they can generate an issues email to notify the loan officer of confirmed issues, or generate an audit report documenting the full QC review for the loan file.

When all review work is finished, the reviewer sets the loan status to Completed using Mark as complete in the drop-down menu. If additional documents are needed, the reviewer uploads them through Manage documents in the Loan documents panel.

To re-evaluate a loan, for example after a corrected document is uploaded, the reviewer uses Refresh review in the header, or Restart review in the drop-down menu.