Virtual Mailroom (VMR) uses AI‑driven automation to identify, classify, and extract information from incoming documents. This user guide provides an overview of how Document AI operates in VMR, including automated document classification, data extraction, and the document automation statuses users see during processing.

• AI-Driven Document Classification
• AI-Driven Data Extraction
• Document Automation Statuses
• What's Next?

AI-Driven Document Classification

AI‑driven document classification refers to the automated categorisation of documents based on their content. VMR uses AI models built on Natural Language Processing (NLP) and Machine Learning (ML) to analyse the text, structure, and available metadata within each document.

Based on this analysis, documents are automatically assigned to predefined document type categories. This enables consistent organisation and supports downstream automation such as document mapping, tagging, and publishing.

VMR Supported Model

Documents that follow recognised structures and patterns can be automatically classified by VMR. These documents are analysed and categorised without user intervention, allowing them to progress through automation workflows efficiently.

VMR Unsupported Model

Documents that do not meet classification requirements—such as those with unclear layouts, unsupported formats, or insufficient identifying information—may not be automatically classified. These documents can still be loaded into VMR but may require manual review before they can proceed.

AI-Driven Data Extraction

AI‑driven data extraction involves automatically identifying and extracting relevant data points from documents and matching them to VMR lookups.

VMR uses Optical Character Recognition (OCR) and NLP to locate, read, and retrieve both structured and unstructured data from scanned PDFs. Extracted data may include entity identifiers, dates, document attributes, and other contextual information used for document mapping and automation.

VMR Supported Model

When documents meet data extraction requirements, VMR automatically extracts key data fields and associates them with the appropriate lookups. This enables documents to be enriched with metadata and supports automated publishing to downstream systems.

VMR Unsupported Model

If a document does not support reliable data extraction—due to format limitations, unreadable content, or missing information—automation may be limited. These documents may still be ingested but require user validation or manual input to complete processing.

Document Automation Statuses

As documents move through Virtual Mailroom (VMR), they are assigned document processing statuses that indicate their current position within the document lifecycle. These statuses are designed to provide transparency, set clear expectations, and highlight when user action may be required, enabling users to quickly assess document progress and identify potential issues at an early stage.

Loaded

Status meaning
The document has been received by VMR and queued for processing, but AI classification and data extraction may not yet be complete.

What typically happens

• The document is queued for AI analysis.
• Document classification to its respective category is not yet complete.
• The document is visible in VMR but is not yet ready for publishing.

Common concerns

• Extended time in Loaded may indicate processing delays.
• Unsupported or password‑protected files may not progress automatically requiring manual intervention.

User expectation
No action is required unless the document remains in the Loaded status longer than expected, in which case escalation to the Class Support team may be required.

Classified

Status meaning
The document has been analysed by AI and progressed from the queue to being assigned a relevant document category, with data identification and mapping to VMR lookups applied where supported, and the extracted data transformed into the appropriate tags.

What typically happens

• AI identifies the document category.
• Key attributes are applied to support automation.
• The document moves to READY for publishing.

Common concerns

• Classification may be incorrect for ambiguous or poor‑quality documents.
• Data extraction and mapping may be incomplete in some cases, for example if the loaded document is not supported by the current model.

User expectation
Documents are progressing normally, but review may be required for accuracy.

Ready

Status meaning
Automation around document classification, data extraction and mapping complete and is ready for publishing to downstream system such as Class DMS or SharePoint.

What typically happens

• Metadata and mapping are finalised.
• The document is eligible for publishing.
• No further AI processing is required.

Common concerns

• Perform a spot check on a small sample of documents to validate overall AI accuracy, including the checks listed below.
• Verify that the document has been classified into the correct document category.
• Confirm that key metadata fields (such as entity, document date, and financial year) have been correctly extracted.
• Check that data mapping to VMR lookups (for example, members, investments, or entities) is accurate where applicable.
• Ensure that system‑generated tags are appropriate and complete for the document type.

User expectation

Minimal human intervention is required; however, if any anomalies are identified during user review, the issue should be escalated to the Class Support team for further investigation.

Errors

Status meaning
The AI could not complete processing due to an issue preventing automation.

Common causes

• Unsupported file types
• Unreadable or low‑quality content
• Password protection
• Missing or unextractable data

What typically happens

• AI processing stops.
• The document cannot progress automatically.
• Human intervention is required.

Common concerns

• Errors can delay workflows if not addressed.
• Repeated errors may indicate source or format issues.

User expectation
Human intervention is required before the document can progress from an Errors state to a Ready state for publishing and, in the event of a system anomaly, may require escalation to the Class Support team.

Document processing statuses in VMR are designed to clearly communicate document progress and indicate when user attention is required. Most documents transition seamlessly from Loaded to Ready through AI‑driven processing, while documents in Errors highlight exceptions that require user review. Regular monitoring of these statuses helps ensure efficient processing and minimises delays in downstream workflows.

What's Next?

Document Review and Validation in VMR