Setting Up Capture Configurations

Last updated: 03.09.2025 • Reading time: ~7 minutes

Capture Configurations are a central component of Smart Processing. They define how documents are analyzed and which data is extracted from them. The extracted data forms the basis for consistent, compliant, and secure processes.

There are two methods available for analyzing documents and extracting data:

Content Understanding Capture Method
Document Intelligence Capture Method

Note

Ensure that the required AI Services are set up.

Content Understanding requires the Content Understanding Service.
Document Intelligence requires the Document Intelligence Service.

Important

Capture Configurations are intended only for standard documents. No individual Capture Configurations are available for XML or email documents, as these are processed automatically through predefined scenarios.

Creating a Capture Configuration

Follow these steps to create a new Capture Configuration:

Start in the Smart Processing Setup area.
In the Setups menu, select the entry Capture Configuration.
Click Create to set up a new configuration.
Follow the Wizard to select the method, define fields, and set validation rules.

Overview of Methods

Content Understanding (CU) Capture Method

With the CU method, you define which data should be extracted from a document based on clear field descriptions.

Particularly suitable for varying or complex document structures.
Fields are recognized content-wise (semantically), regardless of layout.
Ideal when the focus is on content rather than layout.

Document Intelligence (DI) Capture Method

The DI method analyzes documents based on layout.

Uses pre-trained models for typical document types – allows for a quick start.
Optionally, custom models can be trained to reliably process individual or complex layouts.
Suitable when the layout is stable or particularly high accuracy is required.

Which Method When?

Scenario	Recommendation	Justification
Many different layouts, focus on content	Content Understanding	Semantic extraction, robust against layout variations
Standardized documents with stable structure	Document Intelligence (prebuilt)	Quick to set up, low configuration effort
Very specific or complex layouts, high precision required	Document Intelligence (Custom Model)	Maximum stability through training, but higher effort
Quick prototyping or short iteration cycles	Content Understanding	Fields can be easily described and tested directly

Recommendation

Start with Content Understanding if you expect many layout variations or if content is the priority. Use Document Intelligence (prebuilt) for a quick start. Only use Custom Models if CU or prebuilt DI do not provide sufficiently accurate results.

How to Create a Content Understanding Analyzer?

If you have chosen the Content Understanding (CU) method, you can directly create and manage a new analyzer in Business Central. A dedicated configuration page is available for this purpose.

You can open this page either:

in the respective Capture Configuration via the action AI Content Understanding Setup, or
directly through the Smart Processing Setup page.

In this overview, you can see all analyzers that already exist in the service. Each analyzer is stored there and automatically synchronized upon changes.

To create a new analyzer, follow these steps:

Click New to create an analyzer.
Assign a unique name to the analyzer.
Define which fields should be extracted at the Header and Line levels.

For each field, the following information is available:

Setting	Description
Field Name	Technical name of the field to be extracted. Often the name alone is sufficient, but a detailed description is recommended for better results.
Field Description	Free text description that explains the context of the field. Phrase it as if you were explaining to a person where and how to find the field.
Value Type	Expected data type of the field. Available options: `String`, `Date`, `Number`, `Integer`.

Analyzer Actions

Action	Purpose
Schema Management → Import Process Template	Imports fields directly from an existing Process Template. Advantage: Field names are identical to the template, ensuring that matching is 100% accurate later. It also guarantees that all relevant fields are captured in the analyzer.
Test Analyze Tool	Used for testing during setup. 1. Upload a test document. 2. Execute the action Run Test Analyze Tool. 3. A temporary analyzer is created, and a capture simulation is started. 4. The results are displayed in a Preview.
Build Analyzer	Creates the final model. Only built analyzers can be used in production.

Editing Analyzer

You also have the option to extend or adjust existing analyzers. Simply enter an analyzer and edit it as usual. You can then test the analyzer again and rebuild it.

Notes on the Build Process:

Building (Build) usually takes only a few seconds.
You can see the current status via the action Refresh List.
Already built analyzers are immediately available for field extraction.

Best Practice

Always use Import Process Template first before manually defining fields. This ensures that the field names match exactly with the template and that matching works flawlessly later.

Frequently Asked Questions

Do I necessarily need both methods? No. Many scenarios can be fully implemented with one method. Mixed scenarios, e.g., DI for standard invoices and CU for special formats, are possible.

What happens if no service is set up or the credentials are invalid? When starting an action, an error message will be displayed, describing which service is missing or which settings need to be checked.

Summary

Capture Configurations control how data is extracted from documents.
Content Understanding is flexible, semantic, and ideal for varying layouts.
Document Intelligence is layout-oriented, quick with prebuilt models, and particularly robust with custom models.
Choose the method based on layout variety, setup effort, and specific use case.