Choosing the project type

After enabling Document Understanding^TM, the first step is to create a new project:

• Classic: full-featured model-building experience. Use this option to consume projects from different tenants or export datasets or models. If you choose this experience, please check the Document Understanding Classic User Guide for all the information related to the classic experience.
• Modern: guided model-building experience that also includes recommendations for optimal model performance and active learning. Refer to the sections below for current limitations. This user guide covers this experience.
  Note: Before you start using Document Understanding modern projects, make sure that you have an API Key generated in Automation Cloud^TM, Admin > Licenses > Consumables > AI Units page. Once generated, allow for up to 20 minutes for changes to be propagated in the system.

Figure 1. Choosing the project type

A second classic option is to develop the Document Understanding project within AI Center™. Currently, this option gives more control over model training and deployment. However, this option is only recommended for advanced users. For more information about AI Center-based projects, check the AI Center - Using AI Center page from the AI Center User Guide.

The newest and most advanced experience is the experience based on modern projects. However, it's worth noting that modern projects currently have some feature gaps compared to classic experiences. These are planned to be addressed in future releases.

Current limitations

Certain features are currently unavailable for modern projects:

• Auto-fine tuning.
• Classification fields.
• Document tags (batches).

Comparing project types

To better understand the advantages and disadvantages of the different project types, check the following table.

Table 1. Comparing project types

	Modern projects	Classic projects	AI Center-based projects
Advantages	Models and datasets can be exported. Faster model development time and guided training experience. Support for Document Understanding Cloud APIs.	Models and datasets can be exported. Support for Document Understanding Cloud APIs.	Models and datasets can be exported. Support for auto-fine tuning. Available on Automation Suite and standalone deployments.
Disadvantages	The following features are not yet available: Auto-fine-tuning. Classification fields. Document tags (batches).	Slower model development time and no model training guidance. Not available on Automation Suite.	Slower model development time and no model training guidance. No support for Document Understanding Cloud APIs.

Projects compatibility matrix

To better understand the compatibility between different methods, activities, and APIs, check the following table.

Table 2. Compatibility matrix

Project type	Deployment method	Activities compatibility	Document Understanding public APIs compatibility	Available ML model versions
Modern projects	Automation CloudTM Automation Suite	DocumentUnderstanding activities package IntelligentOCR activities package	Yes	24.4
Classic projects	Automation CloudTM	DocumentUnderstanding activities package IntelligentOCR activities package	Yes	24.4 and all older model versions
Projects based on AI Center™	Automation CloudTM Automation Suite Standalone	IntelligentOCR activities package	No	24.4 and all older model versions