AI Integration for OpenText RightFax

AI integration for OpenText RightFax focuses on the inbound fax queue and the document processing workflow. The primary integration points are the RightFax FaxUtil API or document capture hot folders, where inbound TIFF or PDF files land. At this stage, an AI service can be triggered via webhook or a watched directory to perform OCR (Optical Character Recognition) and subsequent analysis. The AI layer acts as a pre-processing engine, analyzing the document content to determine its type (e.g., prescription, invoice, claim form, patient referral), extracting key data fields (patient name, PO number, account ID), and assigning metadata before the document proceeds to its next destination in the workflow.

Implementation typically involves a lightweight microservice or serverless function that receives the fax image, calls a vision/LLM API (like Azure Document Intelligence or Google Document AI) for OCR and classification, and then uses an LLM for entity extraction and intent classification. The enriched data—document type, extracted fields, confidence scores—is then injected back into the RightFax workflow. This can be done by updating the fax's index fields (UserData1-10) via the API or by moving the file to a designated routing folder based on its classification. For example, a fax identified as a medical prescription can be automatically routed to a pharmacy queue folder, while an insurance claim form is sent to a claims processing folder, with its data pre-populated into a downstream system.

Rollout should be phased, starting with a single, high-volume document type to validate accuracy and workflow logic. Governance is critical: a human-in-the-loop review queue should be established for low-confidence classifications or exceptions. Audit trails must log the AI's classification decision, extracted data, and the final routing action for compliance. This architecture doesn't replace RightFax; it augments its routing logic with cognitive understanding, turning a manual triage process into an automated, intelligent intake system. For related architectural patterns, see our guide on Intelligent Document Processing in ECM Platforms.

The integration architecture is event-driven, designed to intercept inbound faxes before they become static PDFs in a user's inbox. The core flow begins when a fax is successfully received by the RightFax server. Using RightFax's FaxUtil API or by monitoring the \Fax\Data\Receive directory, a lightweight connector service captures the fax image (typically a TIFF or PDF). This service immediately posts the document to a secure processing queue (e.g., Azure Service Bus, AWS SQS). This decouples the fax reception—which must be highly reliable—from the subsequent AI processing steps, ensuring fax system stability.

A dedicated AI processing service consumes documents from the queue. Its first step is to run the fax image through a high-accuracy OCR engine (like Azure AI Document Intelligence, Google Document AI, or AWS Textract) to extract machine-readable text and layout information. This OCR'd text, along with the original image, is then passed to a configured LLM (e.g., GPT-4, Claude 3) via a secure API call. The LLM executes a pre-defined prompt chain to perform the core business logic: document classification (e.g., 'Invoice', 'Prescription', 'Insurance Claim'), entity extraction (dates, amounts, patient IDs, PO numbers), and sentiment or urgency scoring. The results are structured into a JSON payload containing the extracted data, confidence scores, and the AI's classification decision.

This enriched data payload is then injected back into the business workflow. Using the RightFax COM API or Enterprise Fax Manager, the system can automatically: 1) Apply metadata to the fax within RightFax, populating custom fields with extracted values for search and reporting; 2) Trigger a RightFax workflow to route the fax to a specific user, department, or network folder based on its classified type; or 3) Post the data to a downstream system via a webhook—such as an ERP (SAP, NetSuite) for invoice processing, an EHR for patient intake, or a case management system. All AI decisions and extracted data are logged to a separate audit database with the original fax ID, enabling human review, model performance tracking, and compliance. The final, routed fax and its AI-generated metadata become a searchable, actionable record within the RightFax environment or connected systems, turning a passive document into a structured workflow trigger.

A production AI integration for RightFax must be architected with zero trust principles and data sovereignty at its core. Inbound faxes often contain PII, PHI, or financial data, requiring processing within your designated environment. The recommended pattern uses a secure, event-driven pipeline: RightFax generates an event (e.g., via the RightFax Enterprise Fax Server API or by dropping a TIFF/PDF into a monitored network folder), which triggers a serverless function in your VPC. This function sends only the document image to a private, compliant AI service (like Azure OpenAI with a VNet) for OCR and analysis. All extracted data and classification tags are written back to RightFax via its API, populating custom metadata fields or routing rules, while the original fax and AI outputs are logged to a secure audit trail. This keeps sensitive document content within your controlled infrastructure at all times.

Governance is enforced through role-based access control (RBAC) on the AI system's outputs and human-in-the-loop approvals for high-stakes decisions. For example, an AI model might classify a fax as a "Patient Lab Result" with 95% confidence and extract critical values, but a configured workflow can still require a nurse's review before the fax is routed to the patient's chart in the EHR. Similarly, extracted data from an insurance claim form can be presented to a claims adjuster in a copilot interface within RightFax or a connected case management system for verification before submission. All AI actions—classification, extraction, routing suggestions—are logged with a user or system ID, creating a transparent chain of custody for compliance audits.

A phased rollout mitigates risk and builds trust. Start with a non-critical, high-volume workflow, such as automatically classifying and routing vendor marketing faxes to a generic inbox. This validates the integration's stability and accuracy in a low-risk scenario. Phase two might target a specific department, like Accounts Payable, to extract invoice numbers, dates, and amounts from supplier faxes for pre-population into your ERP, keeping all human reviewers in the loop. The final phase expands to regulated areas like Healthcare or Financial Services, implementing strict confidence thresholds and mandatory review queues for documents containing sensitive data. This iterative approach allows for tuning of AI models, refinement of prompts, and adaptation of user workflows, ensuring the integration delivers reliable, governed value without disrupting critical business operations.