Generative Extraction — Technical Deep Dive

⏱ ~1 hr 25 min📖 Theory + Hands-On🎯 Expert

🎯 Learning objectives

  • Explain the two approaches to generative extraction in UiPath: Generative Classifier & Extractor, and IXP Generative Extraction
  • Identify when to use each approach based on document complexity, table support, and confidence-scoring needs
  • Describe what Part 1 (Generative Classifier and Extractor) and Part 2 (Generative Extractor in IXP) each cover

🔀 Generative Classifier & Extractor vs IXP Generative Extraction — when to use which?

UiPath offers two distinct approaches to generative extraction. Choosing the right one depends on your document complexity, table requirements, and project setup.

Criteria Generative Classifier & Extractor
(DU Activities — Studio)		 Generative Extraction
(IXP Project)
Table / column field extraction ❌ Not supported — column fields are unsupported; workarounds are unreliable for large tables ✅ Supported — handles complex, multi-column tables
Confidence scores ❌ Generative AI does not produce confidence scores ✅ Provides field-level confidence for human-in-the-loop routing
Max prompts per document 50 prompts maximum per extraction call No strict per-call prompt limit; managed at project level
Response word limit 700 words per field response Handles longer, unstructured responses natively
Best for Teams already building DU workflows in Studio who need quick generative field extraction without a separate IXP project Teams handling complex, high-volume documents needing confidence scoring, table extraction, and full monitoring
💡 Rule of Thumb
Use Generative Classifier & Extractor (DU Activities) for quick, activity-based extraction in existing Studio workflows with simple fields. Switch to IXP Generative Extraction when you need table support, confidence-based routing, project-level management, or a full Build → Measure → Publish lifecycle.

📍 Where to start

Part 1 →

Generative Classifier and Extractor

Generative Classifier, Generative Extractor activity, and good practices for prompt-based extraction in Studio.

Part 2 →

Generative Extractor in IXP

Theory, Practical (Build → Measure → Publish), and Best Practices for the IXP Generative Extraction service.