Automations in Document Understanding™

There are several ways in which you can consume Document Understanding^TM capabilities:

• The DocumentUnderstanding.Activities package is available in Studio Web, Studio X, and Studio Desktop and is pre-configured for you either when you create a new automation starting from a file, or if you continue your journey after publishing a project version.
• Using the IntelligentOCR package, which is designed for Windows and Windows Legacy projects, and pre-configured in the Document Understanding process template.
• Using API calls, consuming Document Understanding as a service via the programming language of your choice.

Choosing the right automation

Check the table below to select the optimal automation method that aligns efficiently with your needs and works best for your projects.

Table 1. Choosing automations

	IntelligentOCR activities package	DocumentUnderstanding.Activities package	Document Understanding APIs
Deployment	Automation Cloud™ Automation Suite Standalone	Automation Cloud™	Automation Cloud™
Best suited for	RPA developers	RPA developers Citizen developers	Users with previous programming experience
Integrated development environment (IDE)	Studio Desktop	Studio Desktop Studio X Studio Web	IDE of your choice
Benefits	Flexibility You can mix extraction and classification models, and can also use extractors and classifiers as fallback You can modify the taxonomy and extraction results using RPA code during run-time. Extensible and open framework You can bring your own classifier, extractor, or OCR engine using the respective interfaces. You have full control over the configuration as an RPA developer. Document Understanding Process Template based on REFramework.	Ease of adoption: Easy to use, available on cloud, no setup required for consuming out-of-the-box models. Can be consumed using the Create Automation option in Document Understanding and Marketplace. Suggested by UiPath® AutopilotTM in workflows. Seamlessly integrated with Document Understanding modern projects, isolated configuration in a Document Understanding project, enabling reusability. Relying on Document Understanding cloud APIs, leading to quicker bug fixes. Single input/output object, Document Data.	Not dependent on a particular technology. Can be consumed from cloud or on-premises environments. Executing without the need of a robot.
Drawbacks	High learning curve Complex configuration, reducing reusability Passing explicit arguments from one activity to the other repeatedly: Taxonomy Document Object Model Text Classification reults Extraction results	Compared to IntelligentOCR, there are some missing features, which are planned to be added: Splitting Business Rules Training (fine-tuning models) On-premises support Support for multiple extraction methods per document type	Requires previous programming experience to be adopted. Does not provide access to other platform capabilities.
Use case	Use if you have an existing automation relying on it for which adopting DocumentUnderstanding.Activities is either impossible (functionality currently not available), or costly.	Use this package if you are just getting started with Document Understanding. Use for new automations relying on: Modern projects Generative capabilities Out-of-the-box specialized models	Use API calls if you want to integrate our services seamlessly into your applications and systems. Use APIs to integrate Document Understanding in microservices architectures which are integrated as software as a service (SaaS).