Prodigy - OpenText Core Capture Services Integration and Automation

Common Integration Use Cases Between Prodigy and OpenText Core Capture Services

Prodigy and OpenText Core Capture Services complement each other well in organizations that need to turn captured documents into high-quality training data for machine learning models. OpenText Core Capture Services handles the intake, classification, and extraction of information from business documents, while Prodigy supports efficient human-in-the-loop annotation and model training. Together, they can improve document automation, reduce manual review effort, and accelerate AI model development.

1. Training document classification models from captured mailroom and intake content

Data flow: OpenText Core Capture Services to Prodigy

Captured documents from digital mailroom, claims intake, or customer onboarding can be exported from OpenText Core Capture Services into Prodigy for manual labeling. Teams can annotate document types such as invoices, purchase orders, contracts, identity documents, or correspondence to create training datasets for custom classification models.

• Improves accuracy of document routing and auto-classification
• Reduces misfiled or manually triaged documents
• Supports continuous model improvement as new document types appear

2. Building extraction models for fields that are difficult to capture reliably

Data flow: OpenText Core Capture Services to Prodigy

When OpenText Core Capture Services extracts key fields such as invoice line items, vendor names, policy numbers, or customer identifiers with low confidence, those documents can be sent to Prodigy for expert annotation. Data scientists can use the labeled examples to train or refine custom extraction models for specific document formats and edge cases.

• Reduces exception handling in accounts payable and onboarding workflows
• Improves extraction quality for nonstandard or variable templates
• Creates a feedback loop for hard-to-parse documents

3. Creating gold-standard datasets for OCR and document understanding model validation

Data flow: OpenText Core Capture Services to Prodigy

Organizations can use documents processed by OpenText Core Capture Services as source material for Prodigy annotation projects to create gold-standard datasets. These datasets can validate OCR accuracy, field extraction performance, and document classification models before deployment into production workflows.

• Provides a controlled benchmark for model testing
• Helps compliance and operations teams verify automation quality
• Supports repeatable model evaluation across document categories

4. Human review of low-confidence capture results to improve automation rules

Data flow: OpenText Core Capture Services to Prodigy and back to OpenText Core Capture Services

Documents flagged by OpenText Core Capture Services for low confidence or exception handling can be routed into Prodigy for annotation by subject matter experts. The corrected labels and extracted values can then be used to retrain models or update capture rules in OpenText Core Capture Services, improving straight-through processing rates over time.

• Targets human effort only where automation is uncertain
• Increases first-pass processing accuracy
• Creates a measurable improvement cycle for capture operations

5. Training NLP models on correspondence and unstructured business communications

Data flow: OpenText Core Capture Services to Prodigy

OpenText Core Capture Services can ingest unstructured correspondence such as emails, letters, complaints, and customer requests. These documents can be exported to Prodigy for text annotation tasks such as intent classification, entity tagging, or topic labeling. The resulting datasets can support NLP models for case routing, sentiment analysis, or automated response workflows.

• Improves handling of unstructured customer communications
• Speeds up triage in service, claims, and legal operations
• Enables more accurate downstream text analytics

6. Accelerating customer onboarding document intelligence

Data flow: OpenText Core Capture Services to Prodigy

In customer onboarding programs, OpenText Core Capture Services can capture identity documents, application forms, and supporting paperwork. Prodigy can then be used to label document elements, signatures, entity fields, and document completeness indicators to train models that detect missing information or classify onboarding packets by readiness.

• Reduces onboarding delays caused by incomplete submissions
• Improves document completeness checks
• Supports faster review by operations and compliance teams

7. Improving invoice and AP exception handling with labeled edge cases

Data flow: OpenText Core Capture Services to Prodigy

Invoices that fail validation in OpenText Core Capture Services, such as those with unusual layouts, poor scan quality, or ambiguous line items, can be sent to Prodigy for annotation. AP teams and data scientists can label the correct vendor, amounts, tax fields, and line-item structures to train models that better handle exceptions.

• Reduces manual invoice correction effort
• Improves automation for nonstandard supplier formats
• Helps AP teams process more invoices without increasing headcount

8. Continuous model improvement using production capture feedback

Data flow: Bi-directional

As OpenText Core Capture Services processes live documents, low-confidence classifications, extraction errors, and user corrections can be fed into Prodigy on a recurring basis. Prodigy can then be used to label the new examples and generate updated training data for periodic model retraining, creating a closed-loop improvement process.

• Keeps models aligned with changing document formats and business rules
• Supports ongoing operational learning without large rework cycles
• Improves long-term ROI from capture automation initiatives

Overall, integrating Prodigy with OpenText Core Capture Services is most valuable when organizations want to move beyond basic document capture and build adaptive AI models that improve classification, extraction, and exception handling over time.