Hello
For those of you who use document understanding to read pdfs, what way do you usually use it?
directly from the studio using the packages/template, or from the orchestator document understanding module (classic or modern?)?
what is the difference between the two? are they to be used together or different alternatives
thank you:)
Depends on what is your requirement
based on that we choose structured,unstructured or comms mining
and everything works best with its own set of activities
cheers
Here is quick difference between both the approaches:
| Category | Classic Document Understanding | Modern Document Understanding |
|---|---|---|
| Release context | Older DU implementation, existing since early DU versions | New, streamlined cloud-first DU experience |
| Where it lives | Orchestrator → Tenant > Document Understanding (Classic) | Orchestrator → Automation Cloud > Modern DU |
| Best for | Developers who want manual control over pipelines | Citizen developers + enterprise teams needing simplified DU |
| Taxonomy management | Uploaded manually (taxonomy.json) | Centralized cloud taxonomy editable in UI |
| Pipeline control | Fully customizable but manual | Predefined pipeline stages with configuration |
| AI model usage | Requires explicit linking of classifiers/extractors | Built-in prebuilt AI models accessible instantly |
| Learning / retraining | Manual retraining workflows needed | Auto-learning built-in, seamless continual improvement |
| Human validation | Validation Station (desktop) or Action Center | Unified cloud-based validation in Action Center |
| Document processing | Queue-based processes; Studio workflow required | Direct document processing pipelines from cloud portal |
| Integration with Studio | Deep integration, but more dev effort | Lightweight — Studio can call Modern DU pipelines |
| Flexibility | Extremely high — every step can be customized | Medium — optimized for simplicity, not custom pipelines |
| Setup time | Long — taxonomy, pipeline, extractors must be set manually | Very fast — click-based configuration |
| Ideal automation pattern | Complex enterprise workflows with special logic | High-volume standardized document types |
| Supported extractors | All extractors (ML, Regex, Form, Intelligent Keyword) | Primarily AI Extractor + prebuilt ML models |
| Deployment complexity | High | Low |
| Use cases | Non-standard documents, custom extractors, advanced routing | Invoices, receipts, POs, IDs, contracts, onboarding documents |
| Recommended by UiPath | Conserved for legacy/advanced scenarios | Future-facing, main DU experience |
I will suggest to follow this structured learning path for more clearer understanding.