The Classification Optimizer Core Agent analyzes usage and detection data of Sensitive Information Types (SITs) in Microsoft Purview and their co-occurrence patterns. It identifies patterns, correlations, coverage gaps, redundant or overlapping classifiers, and generates prioritized recommendations for:

• New composite/combined SIT patterns
• Adjustments to existing SIT parameters (e.g., confidence, instances, thresholds)
• Candidates for Trainable Classifiers for free-form or context-dependent patterns
• Policy optimizations (groupings, scoping, rule tuning)

The agent focuses on delivering an actionable optimization plan that reduces false positives, improves precision/recall of data classification, and surfaces strategic classifier gaps tied to business or regulatory driver