Prodigy - OpenText Decision Service Integration and Automation

Common Integration Use Cases Between Prodigy and OpenText Decision Service

1. Rule-Guided Data Labeling Prioritization

Data flow: OpenText Decision Service ? Prodigy

Use OpenText Decision Service to apply business rules that determine which records, images, or text samples should be prioritized for annotation in Prodigy. For example, claims with high fraud risk, customer complaints from regulated markets, or product images from new suppliers can be routed first to labeling queues.

Business value: Ensures annotation effort is focused on the most business-critical data, reducing wasted labeling time and accelerating model improvement where it matters most.

2. Policy-Based Labeling Workflow Routing

Data flow: Prodigy ? OpenText Decision Service ? workflow or case system

After annotators complete labeling in Prodigy, the labeled output can be evaluated by OpenText Decision Service to determine the next workflow step. For instance, records with low-confidence labels, conflicting annotations, or sensitive content can be routed for human review, legal approval, or escalation to a specialist team.

Business value: Improves quality control and governance by ensuring that exceptions follow defined business policies instead of ad hoc manual handling.

3. Dynamic Labeling Rules Based on Operational Policy Changes

Data flow: OpenText Decision Service ? Prodigy

When business rules change, OpenText Decision Service can provide updated decision criteria that Prodigy uses to adjust annotation guidelines. For example, a bank may update fraud categories, or a healthcare organization may revise document classification rules based on new compliance requirements.

Business value: Keeps labeling operations aligned with current policy without requiring code changes or rework in downstream AI training pipelines.

4. Exception Detection for Active Learning Samples

Data flow: Prodigy ? OpenText Decision Service

Prodigy can surface uncertain or ambiguous samples identified through active learning, and OpenText Decision Service can apply rules to decide whether those samples should be auto-accepted, sent to senior reviewers, or excluded from training. This is especially useful for regulated use cases such as insurance claims, financial documents, or customer communications.

Business value: Reduces annotation risk by applying consistent decision logic to edge cases while preserving the speed benefits of active learning.

5. Compliance-Driven Dataset Segmentation

Data flow: OpenText Decision Service ? Prodigy

OpenText Decision Service can classify incoming data into compliance categories before it reaches Prodigy, such as public, confidential, restricted, or personally identifiable information. Prodigy can then route each category to the appropriate annotation team, apply masking rules, or block certain data from non-authorized users.

Business value: Supports data governance and privacy controls while enabling AI teams to work with sensitive data more safely and efficiently.

6. Annotation Quality Threshold Enforcement

Data flow: Prodigy ? OpenText Decision Service

Prodigy annotation results can be passed to OpenText Decision Service to enforce quality thresholds before data is released to model training. For example, if inter-annotator agreement falls below a defined threshold, the dataset can be held for rework or escalated to a subject matter expert.

Business value: Improves training data reliability and reduces the risk of deploying models trained on inconsistent or low-quality labels.

7. Business Rule Validation for AI Training Data Approval

Data flow: Prodigy ? OpenText Decision Service ? MLOps or data repository

Once data is labeled in Prodigy, OpenText Decision Service can validate whether the dataset meets business criteria before it is approved for model training. Examples include confirming that required label classes are present, ensuring region-specific samples are included, or verifying that prohibited content has been removed.

Business value: Creates a controlled approval gate between annotation and model training, improving consistency and reducing rework in AI delivery pipelines.

8. Feedback Loop for Rule Refinement from Annotation Outcomes

Data flow: Prodigy ? OpenText Decision Service

Patterns discovered during annotation in Prodigy, such as frequent label ambiguity or recurring exception types, can be analyzed and used to refine business rules in OpenText Decision Service. This helps operations and compliance teams update decision logic based on real-world data behavior.

Business value: Enables continuous improvement of both labeling standards and decision policies, strengthening alignment between AI model development and business operations.